Hum vs. anthr
Other vs. anthr
The results obtained with respondent subgroups based on total number of authored peer reviewed publications and total number of authored popular science publications are not shown, because they were not associated with significantly different ( p < .05) means in any comparisons.
*** p < .001; ** p < .01; * p < .05.
Overall scientific experience (as measured with the number of scientific publications authored) had no effect on the scores given to either the dryland or the water‐related hypotheses (Table 2 ). However, the more knowledge the respondents had on human evolution specifically (self‐assessed familiarity with the hypotheses, number of scientific publications on human evolution or experience in teaching human evolution), the lower the scores they gave to the water‐related hypotheses. Among biologists, those who knew more about human evolution were more critical than the less knowledgeable ones, and (paleo)anthropologists were more critical than human biologists with the same self‐assessed knowledge level.
When the dryland hypotheses were split into two groups depending on whether they were based on behavioral arguments or environmental adaptation, both groups obtained rather similar results. The main difference was that the behavioral hypotheses received somewhat higher average credibility scores, which reflects the fact that 10 of the 12 most popular hypotheses were based on behavior (on the other hand, so were the four least popular hypotheses).
To visualize the differences in opinion among the (paleo)anthropologists and representatives of other fields, we repeated the ordination of the hypotheses for each of the four respondent groups separately. In accordance with the fact that most respondents were biologists, the ordination based on the biologists’ data only (Figure 2 c) was very similar to the ordination based on all respondents (Figure 2 a). The ordination based on (paleo)anthropologists’ views (Figure 2 b) differed especially in relation to the hypotheses for bipedalism: Hypotheses that explained bipedalism by foraging, tool use, or carrying were very far removed from the main cloud and toward the opposite side than the water‐related hypotheses. In addition, the average credibility scores given to the water‐related hypotheses were among the lowest of any hypotheses. This contrasted with the situation in the ordination based on human biologists’ data (Figure 2 d), in which the water‐based hypotheses had intermediate credibility scores.
The hypotheses differed clearly from each other in the frequencies of different credibility scores, but there were some similarities in the overall pattern among those six traits for which three or more hypotheses were evaluated (Figure 4 ). None of the hypotheses received the “very likely” score from more than 46% of the respondents, but most traits had at least one hypothesis that was considered “very likely” by more than 23% and likely (either “very likely” or “moderately likely”) by 72%–90%. Many of the intermediately popular hypotheses divided the respondents rather evenly between those who found them likely and those who found them unlikely (the latter referring to the scores “very unlikely” and “moderately unlikely” combined).
Credibility scores given by survey respondents to hypotheses that aim to explain the evolutionary origin of specific human traits. The hypotheses are sorted in order of decreasing popularity as estimated by the percentage of respondents who scored them likely (i.e., either “very likely” or “moderately likely”). Descriptions of the hypotheses as they were given in the survey are shown in Table 1
A causal relationship between articulate speech and descended larynx was accepted by most respondents, but there was no consensus on the direction of the causality. That the larynx descended because this was required by articulate speech was found likely by 84% and very likely by 43%. At the same time, that the evolution of speech was triggered by the descended larynx was found likely by 61% and very likely by 18%. In fact, 36% of the respondents scored both directions as equally likely.
Traits in the category “other” had only one explanatory hypothesis each in the survey, and this was water‐related. All of these hypotheses received many more “very unlikely” than “very likely” scores. However, four hypotheses (that baby swimming, profuse sweating, diving ability, and magnitude of diving reflex evolved as adaptations to a semi‐aquatic way of life) received so many “moderately likely” scores that the percentage of respondents who found them likely was slightly larger than the percentage who found them unlikely (Figure 4 ).
Details on how the hypotheses were scored by respondents representing different fields of expertise are shown in Figure 5 . In accordance with the statistical test results, most hypotheses received rather similar scores from respondents of all fields of expertise. However, (paleo)anthropologists were clearly more critical than representatives of the other fields in relation to several hypotheses, including: that nakedness evolved to avoid ectoparasites, that the big brain evolved because warfare caused pressure for higher intelligence, and that any traits evolved as adaptations to swimming or diving.
Frequencies of credibility scores given to hypotheses aiming to explain different traits (columns) by respondents of different fields of expertise (rows). In each panel, the answers are, from left to right, “very likely,” moderately likely,” “no opinion,” “moderately unlikely,” and “very unlikely.” Hypotheses that have been included in the aquatic ape hypothesis are shown in shades of blue and green. Those dryland hypotheses for which the opinions of anthropologists and other expertise groups clearly diverged are shown in magenta. The other hypotheses are in shades of brown, with darker colors given to hypotheses that received higher average credibility scores in the survey
There was a lot of variation among the traits in how many of the proposed explanations the respondents found convincing (Figure 6 ). For any one trait, 33%–64% of the respondents did not find any of the proposed hypotheses “very likely,” while 19%–38% found exactly one and 8%–45% more than one. Ten respondents (0.8%) explained that they did not score any of the hypotheses as likely, because they do not believe that humans have evolved at all (most of them explicitly referred to special creation by God).
The number of hypotheses (colors) proposed to explain each human trait (rows) that each respondent found very likely (left panel) or likely (either very likely or moderately likely; right panel). The total number of hypotheses included in the survey is shown after the name of each trait
The survey asked respondents’ opinions on twenty critical arguments that have been presented against the aquatic ape hypothesis. For most arguments, the modal response was “no opinion,” especially among those 43% of the respondents who had never heard of AAH before. Nevertheless, some arguments were clearly more frequently agreed with than others (Figure 7 and Table 3 ). The most widely accepted critique was that not all aquatic mammals have naked skin, so hairlessness cannot be considered an aquatic adaptation. In the other extreme, less than 3% of the respondents fully agreed and less than 12% mostly agreed with the critique that AAH is unscientific or not worthy of attention for the reasons given; in most cases, the number of respondents who strongly disagreed with these critiques was larger than the number who mostly or fully agreed.
The degree to which respondents representing different expertise fields agree with critique presented against the aquatic ape hypothesis. The full description of each point of critique can be found in Table 3
Points of critique presented against the aquatic ape hypothesis (AAH). The abbreviations are used in Figure 7 , and the full text is copied verbatim from the survey
Abbreviation | Critique |
---|---|
Hairy aquatics | Not all aquatic mammals have naked skin, so hairlessness cannot be considered an aquatic adaptation. |
Not parsimonious | AAH is less parsimonious than other proposed hypotheses: It has to explain both how human traits evolved in water, and how they were retained after return to land. |
Unnecessary | AAH is not needed, because all human traits can be explained by terrestrial scenarios. |
Coincidence | Humans may be similar to aquatic mammals in some traits, but this is only a coincidence and has no evolutionary relevance. |
No skeletal adaptations | AAH is not supported by fossil evidence, because this shows no skeletal adaptations to an aquatic environment. |
Determinism | A major problem with AAH is that it is based on extreme environmental determinism. |
Nonaquatic fossils | AAH is contradicted by the fossil record, because this suggests a permanently nonaquatic environment. |
Less consistent | AAH is internally less consistent than other proposed hypotheses. |
Apes swim | According to AAH, humans should swim better than apes and have more streamlined bodies, but they do not. |
Not enough time | There has not been enough time for an aquatic phase. |
Comparative anatomy | AAH is merely an exercise in comparative anatomy, not a scientific hypothesis. |
Conflicts evolution | AAH conflicts with what is known about evolutionary processes in general. |
Timing unknown | AAH lacks credibility, because its proponents do not agree on when and where the supposed aquatic phase took place. |
Simplistic | AAH is too simplistic to be taken seriously. |
Not peer‐reviewed | AAH can be ignored, because it was not published in a peer‐reviewed journal, and because it is mostly discussed in forums other than scientific journals. |
False evidence | AAH lacks credibility, because the evidence presented in its favor is false. |
Not professionals | AAH can be ignored, because its main proponents are not professionals in the field of human evolution. |
Pseudoscience | AAH is pseudoscience comparable to creationism. |
Cannot predict | AAH is unscientific, because it cannot make predictions. |
Feministic | AAH is unscientific, because it has been used in feministic argumentation. |
The main results of our survey can be summarized as follows: (1) There was no general agreement among the respondents on why any of the uniquely human traits have evolved: None of the proposed hypotheses was universally either accepted or rejected. (2) For any individual trait, the percentage of respondents who found none of the hypotheses “very likely” was between >30% (bipedalism) and >65% (nakedness). (3) In general, opinions on the credibility of the hypotheses were independent of a person's background (gender, age, field of expertise, degree of scientific experience), but (paleo)anthropologists were clearly more critical than representatives of other fields. (4) The hypotheses that mention adaptation to swimming or diving as an explanatory factor were found much less credible by (paleo)anthropologists and slightly more credible by human biologists than by biologists and representatives of other fields. (5) Most respondents were critical about the aquatic ape hypothesis (AAH), but only a small minority considered it to be unscientific.
Of course, all conclusions based on the survey data must be considered tentative only, because the response rate was very low, and it is possible that the results are biased. Members of some subgroup might have been more likely to respond than members of some other subgroup, and the average credibility scores given to the different hypotheses by the respondents may not be representative of the opinions of all scientists in the background population. However, it is unlikely that a lack of general agreement on the drivers of trait evolution or such a clear difference in opinion between (paleo)anthropologists and others could have emerged just as a result of biased sampling.
Our results did not reveal a set of explanations that would collectively provide a coherent and popular scenario for the origin of all (or even many) human traits. Indeed, some of the hypotheses that had almost equal and rather high average credibility scores explained the same trait, whereas for other traits, no hypothesis emerged as particularly popular. Against this background, it is interesting that almost half of the respondents fully or mostly agreed with the statement that the aquatic ape hypothesis “is not needed, because all human traits can be explained by terrestrial scenarios”.
The lack of agreement on why humans evolved the traits we have today is very obvious in our results: No hypothesis was universally accepted, and for most traits, there were several almost equally popular alternative hypotheses rather than one that would generally be considered superior to the others. None of the hypotheses received the score “very likely” from more than half of the respondents or obtained an average credibility score higher than 4.26 (of 5). For hairlessness, the most popular hypothesis was thought to be “very likely” by only 16% of the respondents, and its average credibility score (3.48) was closer to 3 (which is the limit between being considered more likely than unlikely) than to 4 (moderately likely). In addition, for only two of the traits (subcutaneous fat layer and descended larynx), the most popular hypothesis was found at least moderately likely by almost all respondents at the same time as the next most popular hypothesis was found clearly less likely. This may partly reflect the fact that fewer alternative hypotheses have been proposed for these traits than for many of the others included in the survey.
Importantly, lack of agreement did not reflect just ignorance on the topic among nonspecialists, because the responses were, in general, very similar between anthropologists and respondents representing other fields of science. In fact, anthropologists were even more skeptical about all hypotheses than representatives of the other fields were. In other words, outsiders were slightly more convinced that the proposed hypotheses are plausible than those who work in the field. Maybe anthropologists (especially paleoanthropologists) are more systematically trained to be wary of just‐so‐stories (explanations of past events and processes backed up by little or no evidence) than students in nearby fields are. It is also possible that outsiders are somewhat less likely to question hypotheses proposed within an unfamiliar field. This could be because they do not feel qualified to do so, or because they have not heard of the debates that draw attention to the weaknesses of the hypotheses.
Our results conform with the widespread belief that professionals in the field of human evolution are more critical toward the aquatic ape hypothesis (AAH) than outsiders are (Langdon, 1997 ; Bender et al., 2012 ; see also nonscientific sources such as Hawks, 2005 ; Moore, 2012 and Wikipedia: Aquatic Ape Hypothesis: Talk). However, this did not seem to be due to overall scientific ignorance, because how respondents assessed the credibility of the hypotheses proposing adaptation to swimming or diving was independent of both their overall scientific experience level and how they assessed the credibility of the other hypotheses. Interestingly, those whose main field of expertise is human biology had the most positive attitudes toward the water‐related hypotheses, giving them an average credibility score that was as much as 0.9 units higher (on a 1–5 scale) than the average score given by (paleo)anthropologists.
The difference in average opinion between (paleo)anthropologists and other scientists can be interpreted in two opposite ways. On the one hand, those who know the field of human evolution best may be best positioned to make a justified evaluation of the validity of the alternative hypotheses. On the other hand, prior knowledge may induce one to reject unconventional hypotheses offhand merely because they challenge the established paradigms of a field (Bender et al., 2012 ; Klayman, 1995 ). Obviously, the two interpretations lead to opposite conclusions on whether or not the critical attitude of the (paleo)anthropologists can be taken as evidence that AAH is flawed. In our survey, a vast majority of the respondents who had an opinion on the issue disagreed with the statement that AAH can be ignored because its main proponents are not professionals in the field of human evolution. This was the case both overall and within each field of expertise separately, although the proportion of respondents who agreed with the statement was higher among (paleo)anthropologists than among representatives of the other fields.
In this context, it is also interesting that the respondents’ assessment of the credibility of the water‐related hypotheses did not depend on the number of scientific papers they had authored. This indicates that established scientists are no more likely to reject or accept these hypotheses than junior scientists are—unless their scientific experience relates directly to the field of human evolution. A vast majority of the respondents disagreed with the critique that AAH is unscientific. Of course, this does not mean that they would consider the explanations proposed by AAH to be correct, and indeed, all the hypotheses related to AAH received relatively low credibility scores (although not as low as the least popular dryland hypotheses).
If, for the sake of argument, we accept the most popular explanation for each trait to be the correct one, a scenario of evolution by internal drive emerges: The large brain evolved because complex social organization required higher intelligence, the subcutaneous fat layer evolved to serve as an energy reserve for the developing brain, articulate speech evolved because there was social pressure for elaborate communication, the larynx descended because this was required by articulate speech, bipedalism evolved to make the use of tools and weapons easier, and nakedness evolved to avoid overheating when hunting. For most traits, the next most popular explanation was not far behind in popularity. Most of these were also based on inherent drivers, but sometimes in the opposite temporal sequence (e.g., articulate speech was triggered by the descended larynx; large brain evolved because it was required by articulate speech). We found this result disturbing, because the overwhelming popularity of hypotheses based on inherent drivers gives the impression that human evolution is generally thought to have been goal‐directed. This would be in conflict with the current understanding (explained in every evolutionary biology textbook) that evolution has no foresight.
Overall, the survey revealed no general agreement among the respondents: None of the proposed hypotheses on why specific uniquely human traits have evolved was universally either accepted or rejected. Nevertheless, identifying and quantifying what is not generally known and agreed upon can be useful in itself, as it may help to focus future research on answering the most important open questions. Clearly, there is still a long way to go before the question “why are humans so different from other primates” has been answered in a comprehensive and generally satisfactory way.
Conflict of interest.
None declared.
HT designed and conducted the survey and led the writing. All authors discussed the results and planned the data analyses together. The R code used to analyze the data and draw the figures was written by MT with contributions from JT.
Acknowledgments.
We thank Carlos Peña for writing the code to extract respondents’ email addresses from the Internet; Mirkka Jones, Kalle Ruokolainen, and Timo Vuorisalo for comments that helped to improve the survey questions; and Jouko Tuomisto for comments on the manuscript.
Tuomisto H, Tuomisto M, Tuomisto JT. How scientists perceive the evolutionary origin of human traits: Results of a survey study . Ecol Evol . 2018; 8 :3518–3533. https://doi.org/10.1002/ece3.3887 [ PMC free article ] [ PubMed ] [ Google Scholar ]
Numbers, Facts and Trends Shaping Your World
Read our research on:
Full Topic List
Publications
Read Our Research On:
Majorities say scientific research on gene editing is a misuse – rather than an appropriate use – of technology. But public acceptance of gene editing for babies depends on how it will be used, and views often differ by age and religion.
A new survey shows the number can vary considerably depending how you ask questions about evolution
Tuesday is the 210th anniversary of Charles Darwin’s birth. Roughly eight-in-ten U.S. adults say humans have evolved over time.
Evolution remains a contentious issue. When asked about it, highly religious Americans’ responses can vary depending on how the question is asked.
Measuring public opinion on evolution has never been an easy task for survey researchers.
Almost 160 years after Charles Darwin publicized his groundbreaking theory on the development of life, Americans are still arguing about evolution. In spite of the fact that evolutionary theory is accepted by all but a small number of scientists, it continues to be rejected by many Americans.
Human enhancement may be just around the corner. How do Americans view these emerging technologies that may one day enhance our human capabilities?
Americans are more worried than enthusiastic about using gene editing, brain chip implants and synthetic blood to change human capabilities
A majority of the public says science and religion often conflict, but fewer say science conflicts with their own beliefs. And highly religious Americans are less likely than others to see conflict between faith and science.
The bulk of the analysis in this report stems from a Pew Research Center survey conducted by telephone with a national sample of adults (18 years of age or older) living in all 50 U.S. states and the District of Columbia. The results are based on 2,002 interviews (801 respondents were interviewed on a landline […]
901 E St. NW, Suite 300 Washington, DC 20004 USA (+1) 202-419-4300 | Main (+1) 202-857-8562 | Fax (+1) 202-419-4372 | Media Inquiries
ABOUT PEW RESEARCH CENTER Pew Research Center is a nonpartisan, nonadvocacy fact tank that informs the public about the issues, attitudes and trends shaping the world. It does not take policy positions. The Center conducts public opinion polling, demographic research, computational social science research and other data-driven research. Pew Research Center is a subsidiary of The Pew Charitable Trusts , its primary funder.
© 2024 Pew Research Center
Advances in machine learning and deep learning research are reshaping our technology. Machine learning and deep learning have accomplished various astounding feats, and key research articles have resulted in technical advances used by billions of people. The research in this sector is advancing at a breakneck pace and assisting you to keep up. Here is a collection of the most important scientific study papers in machine learning.
The authors of this work examined why ACGAN training becomes unstable as the number of classes in the dataset grows. The researchers revealed that the unstable training occurs due to a gradient explosion problem caused by the unboundedness of the input feature vectors and the classifier’s poor classification capabilities during the early training stage. The researchers presented the Data-to-Data Cross-Entropy loss (D2D-CE) and the Rebooted Auxiliary Classifier Generative Adversarial Network to alleviate the instability and reinforce ACGAN (ReACGAN). Additionally, extensive tests of ReACGAN demonstrate that it is resistant to hyperparameter selection and is compatible with a variety of architectures and differentiable augmentations.
This article is ranked #1 on CIFAR-10 for Conditional Image Generation.
For the research paper, read here .
For code, see here .
The authors presented a straightforward and computationally fast unsupervised strategy for learning dense spacetime representations from unlabeled films in this study. The approach demonstrates rapid convergence of training and a high degree of data efficiency. Furthermore, the researchers obtain VOS accuracy superior to previous results despite employing a fraction of the previously necessary training data. The researchers acknowledge that the research findings may be utilised maliciously, such as for unlawful surveillance, and that they are excited to investigate how this skill might be used to better learn a broader spectrum of invariances by exploiting larger temporal windows in movies with complex (ego-)motion, which is more prone to disocclusions.
This study is ranked #1 on DAVIS 2017 for Unsupervised Video Object Segmentation (val).
The authors offer an atlas-based technique for producing unsupervised temporally consistent surface reconstructions by requiring a point on the canonical shape representation to translate to metrically consistent 3D locations on the reconstructed surfaces. Finally, the researchers envisage a plethora of potential applications for the method. For example, by substituting an image-based loss for the Chamfer distance, one may apply the method to RGB video sequences, which the researchers feel will spur development in video-based 3D reconstruction.
This article is ranked #1 on ANIM in the category of Surface Reconstruction.
The researchers propose a revolutionary interactive architecture called EdgeFlow that uses user interaction data without resorting to post-processing or iterative optimisation. The suggested technique achieves state-of-the-art performance on common benchmarks due to its coarse-to-fine network design. Additionally, the researchers create an effective interactive segmentation tool that enables the user to improve the segmentation result through flexible options incrementally.
This paper is ranked #1 on Interactive Segmentation on PASCAL VOC
The authors of this work examined whether it is possible to transfer the success of task-agnostic web-scale pre-training in natural language processing to another domain. The findings indicate that adopting this formula resulted in the emergence of similar behaviours in the field of computer vision, and the authors examine the social ramifications of this line of research. CLIP models learn to accomplish a range of tasks during pre-training to optimise their training objective. Using natural language prompting, CLIP can then use this task learning to enable zero-shot transfer to many existing datasets. When applied at a large scale, this technique can compete with task-specific supervised models, while there is still much space for improvement.
This research is ranked #1 on Zero-Shot Transfer Image Classification on SUN
The researchers in this article conduct a thorough examination of the features of convolutions and transformers, resulting in a principled approach for combining them into a new family of models dubbed CoAtNet. Extensive experiments demonstrate that CoAtNet combines the advantages of ConvNets and Transformers, achieving state-of-the-art performance across a range of data sizes and compute budgets. Take note that this article is currently concentrating on ImageNet classification for model construction. However, the researchers believe their approach is relevant to a broader range of applications, such as object detection and semantic segmentation.
This paper is ranked #1 on Image Classification on ImageNet (using extra training data).
The authors of this article suggest the SwinIR image restoration model, which is based on the Swin Transformer . The model comprises three modules: shallow feature extraction, deep feature extraction, and human-recognition reconstruction. For deep feature extraction, the researchers employ a stack of residual Swin Transformer blocks (RSTB), each formed of Swin Transformer layers, a convolution layer, and a residual connection.
This research article is ranked #1 on Image Super-Resolution on Manga109 – 4x upscaling.
Ways to incorporate historical data are still unclear: initialising reward estimates with historical samples can suffer from bogus and imbalanced data coverage, leading to computational and storage issues—particularly in continuous action spaces. The paper addresses the obstacles by proposing ‘Artificial Replay’, an algorithm to incorporate historical data into any arbitrary base bandit algorithm.
Read the full paper here .
Author(s) – Sean R. Sinclair et al.
The paper proposes an algorithm in which the meta-learner teaches itself to overcome the meta-optimisation challenge. The algorithm focuses on meta-learning with gradients, which guarantees performance improvements. Furthermore, the paper also looks at how bootstrapping opens up possibilities.
Read the full paper here .
Author(s) – Sebastian Flennerhag et al.
The research describes the LaMDA system which caused chaos in AI this summer when a former Google engineer claimed that it had shown signs of sentience. LaMDA is a family of large language models for dialogue applications based on Transformer architecture. The interesting feature of the model is its fine-tuning with human-annotated data and the possibility of consulting external sources. This is a very interesting model family, which we might encounter in many applications we use daily.
Author(s) – Yujia Li et al.
Systems can help programmers become more productive. The following research addresses the problems with incorporating innovations in AI into these systems. AlphaCode is a system that creates solutions for problems that require deeper reasoning.
Author(s) – Tian Dong et al.
The paper focuses on Privacy Preserving Machine Learning, specifically deducting the leakage of sensitive data in machine learning. It puts forth one of the first propositions of using dataset condensation techniques to preserve the data efficiency during model training and furnish membership privacy.
Author(s) – Léo Grinsztajn, Edouard Oyallon and Gaël Varoquaux
The research answers why deep learning models still find it hard to compete on tabular data compared to tree-based models. It is shown that MLP-like architectures are more sensitive to uninformative features in data compared to their tree-based counterparts.
Author(s) – Samuel Daulton et al.
The paper proposes ‘MORBO’, a scalable method for multiple-objective BO as it performs better than that of high-dimensional search spaces. MORBO significantly improves the sample efficiency and, where existing BO algorithms fail, MORBO provides improved sample efficiencies over the current approach.
Author(s) – Yann LeCun
The research offers a vision about how to progress towards general AI. The study combines several concepts: a configurable predictive world model, behaviour driven through intrinsic motivation, and hierarchical joint embedding architectures trained with self-supervised
learning.
Author(s) – Shreshth Tuli, Giuliano Casale and Nicholas R. Jennings
This is a specialised paper applying transformer architecture to the problem of unsupervised anomaly detection in multivariate time series. Many architectures which were successful in other fields are, at some point, also being applied to time series. The research shows improved performance on some known data sets.
Author(s) – Zhiqi Bu et al.
In the paper, researchers study the problem of differentially private (DP) fine-tuning of large pre-trained models—a recent privacy-preserving approach suitable for solving downstream tasks with sensitive data. Existing work has demonstrated that high accuracy is possible under strong privacy constraints yet requires significant computational overhead or modifications to the network architecture.
Usually, increasing model size when pretraining natural language representations often result in improved performance on downstream tasks, but the training times become longer. To address these problems, the authors in their work presented two parameter-reduction techniques to lower memory consumption and increase the training speed of BERT. The authors also used a self-supervised loss that focuses on modelling inter-sentence coherence and consistently helped downstream tasks with multi-sentence inputs. According to results, this model established new state-of-the-art results on the GLUE, RACE, and squad benchmarks while having fewer parameters compared to BERT-large.
Check the paper here .
Microsoft Research, along with the University of Washington and the University of California, in this paper, introduced a model-agnostic and task agnostic methodology for testing NLP models known as CheckList. This is also the winner of the best paper award at the ACL conference this year. It included a matrix of general linguistic capabilities and test types that facilitate comprehensive test ideation, as well as a software tool to generate a large and diverse number of test cases quickly.
Linformer is a Transformer architecture for tackling the self-attention bottleneck in Transformers. It reduces self-attention to an O(n) operation in both space- and time complexity. It is a new self-attention mechanism which allows the researchers to compute the contextual mapping in linear time and memory complexity with respect to the sequence length.
Read more about the paper here .
Plug and Play Language Models ( PPLM ) are a combination of pre-trained language models with one or more simple attribute classifiers. This, in turn, assists in text generation without any further training. According to the authors, model samples demonstrated control over sentiment styles, and extensive automated and human-annotated evaluations showed attribute alignment and fluency.
The researchers at Google, in this paper , introduced Reformer. This work showcased that the architecture of a Transformer can be executed efficiently on long sequences and with small memory. The authors believe that the ability to handle long sequences opens the way for the use of the Reformer on many generative tasks. In addition to generating very long coherent text, the Reformer can bring the power of Transformer models to other domains like time-series forecasting, music, image and video generation.
The irony here is that one of the popular language models, Transformers have been made to do computer vision tasks. In this paper , the authors claimed that the vision transformer could go toe-to-toe with the state-of-the-art models on image recognition benchmarks, reaching accuracies as high as 88.36% on ImageNet and 94.55% on CIFAR-100. For this, the vision transformer receives input as a one-dimensional sequence of token embeddings. The image is then reshaped into a sequence of flattened 2D patches. The transformers in this work use constant widths through all of its layers.
Winner of the CVPR best paper award, in this work, the authors proposed a method to learn 3D deformable object categories from raw single-view images, without external supervision. This method uses an autoencoder that factored each input image into depth, albedo, viewpoint and illumination. The authors showcased that reasoning about illumination can be used to exploit the underlying object symmetry even if the appearance is not symmetric due to shading.
In this paper, OpenAI researchers examined whether similar models can learn useful representations for images. For this, the researchers trained a sequence Transformer to auto-regressively predict pixels, without incorporating knowledge of the 2D input structure. Despite training on low-resolution ImageNet without labels, the researchers found that a GPT-2 scale model learns strong image representations as measured by linear probing, fine-tuning, and low-data classification. On CIFAR-10, it achieved 96.3% accuracy with a linear probe, outperforming a supervised Wide ResNet, and 99.0% accuracy with full fine-tuning and matching the top supervised pre-trained models. An even larger model, trained on a mixture of ImageNet and web images, is competitive with self-supervised benchmarks on ImageNet, achieving 72.0% top-1 accuracy on a linear probe of their features.
In this paper, the authors provided a high-level introduction to deep RL , discussed some of its initial applications to neuroscience, and surveyed its wider implications for research on brain and behaviour and concluded with a list of opportunities for next-stage research. Although DeepRL seems to be promising, the authors wrote that it is still a work in progress and its implications in neuroscience should be looked at as a great opportunity. For instance, deep RL provides an agent-based framework for studying the way that reward shapes representation, and how representation, in turn, shapes learning and decision making — two issues which together span a large swath of what is most central to neuroscience.
Why humans doing certain things are often linked to dopamine , a hormone that acts as the reward system (think: the likes on your Instagram page). So, keeping this fact in hindsight, DeepMind with the help of Harvard labs, analysed dopamine cells in mice and recorded how the mice received rewards while they learned a task. They then checked these recordings for consistency in the activity of the dopamine neurons with standard temporal difference algorithms. This paper proposed an account of dopamine-based reinforcement learning inspired by recent artificial intelligence research on distributional reinforcement learning. The authors hypothesised that the brain represents possible future rewards not as a single mean but as a probability distribution, effectively representing multiple future outcomes simultaneously and in parallel.
In this paper, the authors bridged natural language processing (NLP) and reinforcement learning (RL). They examined both recurrent LSTM models and large-scale Transformer models for NLP and discrete-action space tasks for RL. The results suggested that the lottery ticket hypothesis is not restricted to supervised learning of natural images, but rather represents a broader phenomenon in deep neural networks.
In this paper, the authors explored if the reward function itself can be a good locus of learned knowledge. They proposed a scalable framework for learning useful intrinsic reward functions across multiple lifetimes of experience and showed that it is feasible to learn and capture knowledge about long-term exploration and exploitation into a reward function.
The progress of AutoML has largely focused on the architecture of neural networks, where it has relied on sophisticated expert-designed layers as building blocks, or similarly restrictive search spaces. In this paper , the authors showed that AutoML could go further with AutoML Zero, that automatically discovers complete machine learning algorithms just using basic mathematical operations as building blocks. The researchers demonstrated this by introducing a novel framework that significantly reduced human bias through a generic search space.
Batch normalization is an essential component of meta-learning pipelines. However, there are several challenges. So, in this paper, the authors evaluated a range of approaches to batch normalization for meta-learning scenarios and developed a novel approach — TaskNorm. Experiments demonstrated that the choice of batch normalization has a dramatic effect on both classification accuracy and training time for both gradient-based and gradient-free meta-learning approaches. The TaskNorm has been found to be consistently improving the performance.
Meta-learning algorithms need meta-training tasks to be mutually exclusive, such that no single model can solve all of the tasks at once. In this paper, the authors designed a meta-regularisation objective using information theory that successfully uses data from non-mutually-exclusive tasks to efficiently adapt to novel tasks.
Model Agnostic Meta-Learning (MAML) consists of optimisation loops, from which the inner loop can efficiently learn new tasks. In this paper, the authors demonstrated that feature reuse is the dominant factor and led to ANIL (Almost No Inner Loop) algorithm — a simplification of MAML where the inner loop is removed for all but the (task-specific) head of the underlying neural network.
This paper proposed attempts to reinterpret a standard discriminative classifier as an energy-based model. In this setting, wrote the authors, the standard class probabilities can be easily computed. They demonstrated that energy-based training of the joint distribution improves calibration, robustness, handout-of-distribution detection while also enabling the proposed model to generate samples rivalling the quality of recent GAN approaches. This work improves upon the recently proposed techniques for scaling up the training of energy-based models. It has also been the first to achieve performance rivalling the state-of-the-art in both generative and discriminative learning within one hybrid model.
This paper investigated the commonly assumed notion that neural networks cannot be recovered from its outputs, as they depend on its parameters in a highly nonlinear way. The authors claimed that by observing only its output, one could identify the architecture, weights, and biases of an unknown deep ReLU network. By dissecting the set of region boundaries into components associated with particular neurons, the researchers showed that it is possible to recover the weights of neurons and their arrangement within the network.
Authors: Suyash Mahajan, Salma Shaikh, Jash Vora, Gunjan Kandhari, Rutuja Pawar,
Abstract: The paper embark on predicting the outcomes of Indian Premier League (IPL) cricket match using a supervised learning approach from a team composition perspective. The study suggests that the relative team strength between the competing teams forms a distinctive feature for predicting the winner. Modeling the team strength boils down to modeling individual player‘s batting and bowling performances, forming the basis of our approach.
Research Methodology: In this paper, two methodologies have been used. MySQL database is used for storing data whereas Java for the GUI. The algorithm used is Clustering Algorithm for prediction. The steps followed are as
For the research paper, read here
Author : Roshan Tavhare
Institute : University of Mumbai
Abstract : The main idea behind this project is to develop a nonintrusive system which can detect fatigue of any human and can issue a timely warning. Drivers who do not take regular breaks when driving long distances run a high risk of becoming drowsy a state which they often fail to recognize early enough.
Research Methodology : A training set of labeled facial landmarks on an image. These images are manually labeled, specifying specific (x, y) -coordinates of regions surrounding each facial structure.
Authors: Chinmaya Mishra Praveen Kumar and Reddy Kumar Moda, Syed Saqib Bukhari and Andreas Dengel
Institute: German Research Center for Artificial Intelligence (DFKI), Kaiserslautern, Germany
Abstract: In this paper, the researchers explore various text data augmentation techniques in text space and word embedding space. They studied the effect of various augmented datasets on the efficiency of different deep learning models for relation classification in text.
Research Methodology: The researchers implemented five text data augmentation techniques (Similar word, synonyms, interpolation, extrapolation and random noise method) and explored the ways in which we could preserve the grammatical and the contextual structures of the sentences while generating new sentences automatically using data augmentation techniques.
Author : Prateek Kaushik
Institute : G D Goenka University, Gurugram
Abstract : This research paper described a personalised smart health monitoring device using wireless sensors and the latest technology.
Research Methodology: Machine learning and Deep Learning techniques are discussed which works as a catalyst to improve the performance of any health monitor system such supervised machine learning algorithms, unsupervised machine learning algorithms, auto-encoder, convolutional neural network and restricted boltzmann machine .
Author : A.Pavithra, C.Anandhakumar and V.Nithin Meenashisundharam
Institute : Sree Saraswathi Thyagaraja College,
Abstract : This article we discuss about Big data on IoT and how it is interrelated to each other along with the necessity of implementing Big data with IoT and its benefits, job market
Research Methodology : Machine learning, Deep Learning, and Artificial Intelligence are key technologies that are used to provide value-added applications along with IoT and big data in addition to being used in a stand-alone mod.
Author: Stephen Merity
In this work of art, the Harvard grad author, Stephen “Smerity” Merity, investigated the current state of NLP, the models being used and other alternate approaches. In this process, he tears down the conventional methods from top to bottom, including etymology.
The author also voices the need for a Moore’s Law for machine learning that encourages a minicomputer future while also announcing his plans on rebuilding the codebase from the ground up both as an educational tool for others and as a strong platform for future work in academia and industry.
Authors: Mingxing Tan and Quoc V. Le
In this work, the authors propose a compound scaling method that tells when to increase or decrease depth, height and resolution of a certain network.
Convolutional Neural Networks(CNNs) are at the heart of many machine vision applications.
EfficientNets are believed to superpass state-of-the-art accuracy with up to 10x better efficiency (smaller and faster).
Authors: Mikhail Belkin, Daniel Hsu, Siyuan Ma, Soumik Mandal
In this paper , an attempt has been made to reconcile classical understanding and modern practice within a unified performance curve.
The “double descent” curve overtakes the classic U-shaped bias-variance trade-off curve by showing how increasing model capacity beyond the point of interpolation results in improved performance.
Authors: Jonathan Frankle, Michael Carbin
Neural network pruning techniques can reduce the parameter counts of trained networks by over 90%, decreasing storage requirements and improving computational performance of inference without compromising accuracy.
The authors find that a standard pruning technique naturally uncovers subnetworks whose initializations made them capable of training effectively. Based on these results, they introduce the “lottery ticket hypothesis:”
Authors: Francois Chollet
This work summarizes and critically assesses the definitions of intelligence and evaluation approaches, while making apparent the historical conceptions of intelligence that have implicitly guided them.
The author, also the creator of keras, introduces a formal definition of intelligence based on Algorithmic Information Theory and using this definition, he also proposes a set of guidelines for what a general AI benchmark should look like.
Authors: Sawan Kumar, Sharmistha Jat, Karan Saxena and Partha Talukdar
Word Sense Disambiguation (WSD) is a longstanding but open problem in Natural Language Processing (NLP). Current supervised WSD methods treat senses as discrete labels and also resort to predicting the Most-Frequent-Sense (MFS) for words unseen during training.
The researchers from IISc Bangalore in collaboration with Carnegie Mellon University propose Extended WSD Incorporating Sense Embeddings (EWISE), a supervised model to perform WSD by predicting over a continuous sense embedding space as opposed to a discrete label space.
Authors: Shaojie Bai, J. Zico Kolter and Vladlen Koltun
Motivated by the observation that the hidden layers of many existing deep sequence models converge towards some fixed point, the researchers at Carnegie Mellon University present a new approach to modeling sequential data through deep equilibrium model (DEQ) models.
Using this approach, training and prediction in these networks require only constant memory, regardless of the effective “depth” of the network.
Authors: Robert G, Patricia R, Claudio M, Matthias Bethge, Felix A. W and Wieland B
Convolutional Neural Networks (CNNs) are commonly thought to recognise objects by learning increasingly complex representations of object shapes. The authors in this paper , evaluate CNNs and human observers on images with a texture-shape cue conflict. They show that ImageNet-trained CNNs are strongly biased towards recognising textures rather than shapes, which is in stark contrast to human behavioural evidence.
Authors: Marc G. B , Will D , Robert D , Adrien A T , Pablo S C , Nicolas Le R , Dale S, Tor L, Clare L
The authors propose a new perspective on representation learning in reinforcement learning
based on geometric properties of the space of value functions. This work shows that adversarial value functions exhibit interesting structure, and are good auxiliary tasks when learning a representation of an environment. The authors believe this work to open up the possibility of automatically generating auxiliary tasks in deep reinforcement learning.
Authors: Adam Gaier & David Ha
In this work , the authors explore whether neural network architectures alone, without learning any weight parameters, can encode solutions for a given task. In this paper, they propose a search method for neural network architectures that can already perform a task without any explicit weight training.
Authors: Prajit Ramachandran, Niki P, Ashish Vaswani,Irwan Bello Anselm Levskaya, Jonathon S
In this work, the Google researchers verified that content-based interactions can serve the vision models . The proposed stand-alone local self-attention layer achieves competitive predictive performance on ImageNet classification and COCO object detection tasks while requiring fewer parameters and floating-point operations than the corresponding convolution baselines. Results show that attention is especially effective in the later parts of the network.
Authors: Mario Lucic, Michael Tschannen, Marvin Ritter, Xiaohua Z, Olivier B and Sylvain Gelly
Modern-day models can produce high quality, close to reality when fed with a vast quantity of labelled data. To solve this large data dependency, researchers from Google released this work , to demonstrate how one can benefit from recent work on self- and semi-supervised learning to outperform the state of the art on both unsupervised ImageNet synthesis, as well as in the conditional setting.
The proposed approach is able to match the sample quality of the current state-of-the-art conditional model BigGAN on ImageNet using only 10% of the labels and outperform it using 20% of the labels.
Authors: Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin G, Piyush Sharma and Radu S
The authors present two parameter-reduction techniques to lower memory consumption and increase the training speed of BERT and to address the challenges posed by increasing model size and GPU/TPU memory limitations, longer training times, and unexpected model degradation
As a result, this proposed model establishes new state-of-the-art results on the GLUE, RACE, and SQuAD benchmarks while having fewer parameters compared to BERT-large.
Author: Taesung Park, Ming-Yu Liu, Ting-Chun Wang and Jun-Yan Zhu
Nvidia in collaboration with UC Berkeley and MIT proposed a model which has a spatially-adaptive normalization layer for synthesizing photorealistic images given an input semantic layout.
This model retained visual fidelity and alignment with challenging input layouts while allowing the user to control both semantic and style.
📣 Want to advertise in AIM? Book here
Biggest ai stories, delivered to your inbox every week..
Discover how Cypher 2024 expands to the USA, bridging AI innovation gaps and tackling the challenges of enterprise AI adoption
© Analytics India Magazine Pvt Ltd & AIM Media House LLC 2024
The Belamy, our weekly Newsletter is a rage. Just enter your email below.
IMAGES
VIDEO
COMMENTS
Evolution is the process of heritable change in populations of organisms over multiple generations. Evolutionary biology is the study of this process, which can occur through mechanisms including ...
Abstract. The latest periods of human evolution are a heated topic of debate and have been at the center of paleoanthropological discussions since the beginning of the field. In the last twenty years, new excavations increased the geographic range of paleoanthropological data, new fossil hominins of the last third of the Pleistocene were found ...
María Martinón-Torres is a palaeoanthropologist, director of the National Research Centre on Human Evolution (CENIEH) in Burgos, Spain, and an honorary reader at University College London.
Human evolution refers to the natural. process involved in the evolutionary history of all. members of the human clade (consisting of Homo. and other members of the human tribe, Hominini, after ...
Evolution, the official journal of the Society for the Study of Evolutionà ¢à  à ¢, publishes articles in all areas of evolutionary biology focused on broadening understanding of evolutionary phenomena and processes at all levels of biological organization.
Evolution of ion channels in cetaceans: a natural experiment in the tree of life. Cristóbal Uribe. , Mariana F. Nery. & Juan C. Opazo. Article. 23 July 2024 | Open Access. Earliest evidence of ...
The Evidence for Evolution. Alan R. Rogers. University of Chicago Press, 2011. 128 pp., illus. $18.00 (ISBN 9780226723822 paper). Although scientists view evolution as an indisputable feature of the natural world, most Americans simply do not believe that it occurs, or they reject naturalistic explanations for biotic change.Empirical studies have revealed that students and teachers often know ...
About the journal. JEB is an international, peer-reviewed journal that covers diverse research areas in evolutionary biology. The journal prioritizes publishing significant advances from a broad taxonomic perspective. Find out more.
However, the recent discovery of modern human fossils in Greece and Israel dating to about 210 to 177 ka ago (9, 10) and ancient European genomes show that there were multiple out-of-Africa dispersals in the last 400,000 years, during which early humans and Neanderthals interbred (11, 12). Unlike what happened 60 ka ago (13), the offspring grew ...
Charles Darwin had been thinking about evolution since 1836, and accumulating evidence both that it had occurred and that it was caused primarily (but not exclusively, as he was always careful to emphasize) by natural selection. By 1844, he had the preliminary ideas on evolution by natural selection committed to paper.
Periodically, groups of scientists meet to identify new opportunities in evolutionary biology and associated disciplines (e.g., , -).Rather than set a specific research agenda for the future—clearly the charge of individual investigators—the aim has been to identify new themes and research directions that are already emerging in the field and to focus on the intersection of fundamental ...
Abstract. Evolution is both a fact and a theory. Evolution is widely observable in laboratory and natural populations as they change over time. The fact that we need annual flu vaccines is one ...
Evolution is a scientific theory in biological sciences, which explains the emergence of new varieties of living things in the past and present. Evolution accounts for the conspicuous patterns of similarities and differences among living things over time and across habitats through the action of biological processes such as mutation, natural ...
Evolution refers to change through time as species become modified and diverge to produce multiple descendant species. Evolution and natural selection are often conflated, but evolution is the historical occurrence of change, and natural selection is one mechanism—in most cases the most important—that can cause it.
Abstract. Evolution is both a fact and a theory. Evolution is widely observable in laboratory and natural populations as they change over time. The fact that we need annual flu vaccines is one example of observable evolution. At the same time, evolutionary theory explains more than observations, as the succession on the fossil record.
Clonal evolution is now a central theoretical framework in cancer research. In this Perspective, Laplane and Maley identify challenges to that theory such that some non-evolutionary phenomena in ...
The origin of birds is now one of the best understood major transitions in the history of life. It has emerged as a model case for using a combination of data from fossils, living species, genealogies, and numerical analyses to study how entirely new body plans and behaviors originate, and how prominent living groups achieved their diversity over hundreds of millions of years of evolution 2, 3.
Comprehensive understanding of evolution is essential to full and meaningful engagement with issues facing societies today. Yet this understanding is challenged by lack of acceptance of evolution as well as misconceptions about how evolution works that persist even after student completion of college-level life science courses. Recent research has suggested that active learning strategies, a ...
Learning of evolution by natural selection. The theory of evolution by natural selection is a core feature of biology and centres in many science curricula from around the world (Deniz & Borgerding, Citation 2018).However it is notoriously difficult to teach for various reasons including potential conflict with worldviews and difficulties with understanding the key concepts involved.
From subtle shifts in the genetic makeup of a single population to the entire tree of life, evolution is the process by which life changes from one generation to the next and from one geological epoch to another. The study of evolution encompasses both the historical pattern of evolu-tion—who gave rise to whom, and when, in the tree of life ...
Figure 1. Male and female human figures from the plaque of the Pioneer 10 and 11 spacecrafts. The pictorial message was intended to describe the origin of the probe for potential extraterrestrial life. It shows several typically human traits, such as bipedalism, nakedness, arched nose, large head, and opposable thumbs.
Appendix A: About the Survey. The bulk of the analysis in this report stems from a Pew Research Center survey conducted by telephone with a national sample of adults (18 years of age or older) living in all 50 U.S. states and the District of Columbia. The results are based on 2,002 interviews (801 respondents were interviewed on a landline […]
Click on the title to browse this journal
For the research paper, read here. Smart Health Monitoring and Management Using Internet of Things, Artificial Intelligence with Cloud Based Processing. Author: Prateek Kaushik. Institute: G D Goenka University, Gurugram. Abstract: This research paper described a personalised smart health monitoring device using wireless sensors and the latest ...