Nov 24, 2021

Synthetic Intelligence Index Report 2021

Written by

The AI Index is an independent initiative at Stanford University’s Human-Centered Artificial Intelligence Institute (HAI).As artificial intelligence-powered innovations become ever more prevalent in our lives, the ethical challenges of AI applications are increasingly evident and subject to scrutiny.RESEARCH & DEVELOPMENT [1]source:• The number of AI journal publications grew by 34.5% from 2019 to 2020 —…

The AI Index is an unbiased initiative at Stanford College’s Human-Centered Synthetic Intelligence Institute (HAI).As synthetic intelligence-powered improvements change into ever extra prevalent in our lives, the moral challenges of AI purposes are more and more evident and topic to scrutiny.RESEARCH & DEVELOPMENT [1]supply:• The variety of AI journal publications grew by 34.5% from 2019 to 2020 — a a lot greater proportion progress than from 2018 to 2019 (19.6%).• In each main nation and area, the very best proportion of peer-reviewed AI papers comes from educational establishments. However the second most necessary originators are totally different: In america, corporate-affiliated analysis represents 19.2% of the entire publications, whereas the federal government is the second most necessary in China (15.6%) and the European Union (17.2%).• In 2020, and for the primary time, China surpassed america within the share of AI journal citations on the earth, having briefly overtaken america within the total variety of AI journal publications in 2004 after which retaken the lead in 2017. Nevertheless, america has constantly (and considerably) extra cited AI convention papers than China over the past decade.In response to COVID-19, most main AI conferences befell just about and registered a big enhance in attendance because of this. The variety of attendees throughout 9 conferences virtually doubled in 2020.In simply the final six years, the variety of AI-related publications on arXiv grew by greater than sixfold, from 5,478 in 2015 to 34,736 in 2020. • AI publications represented 3.8% of all peer-reviewed scientific publications worldwide in 2019, up from 1.3% in 20112. TECHNICAL of the whole lot: AI techniques can now compose textual content, audio, and pictures to a sufficiently excessive normal that people have a tough time telling the distinction between artificial and non-synthetic outputs for some constrained purposes of the expertise. It guarantees to generate an amazing vary of downstream purposes of AI for each socially helpful and fewer helpful functions. It is usually inflicting researchers to put money into applied sciences for detecting generative fashions; the DeepFake Detection Problem information signifies how properly computer systems can distinguish between totally different outputs.The industrialization of laptop imaginative and prescient: Pc imaginative and prescient has seen immense progress prior to now decade, primarily as a consequence of the usage of machine studying strategies (particularly deep studying). New information reveals that laptop imaginative and prescient is industrializing: Efficiency is beginning to flatten on a few of the largest benchmarks, suggesting that the neighborhood must develop and agree on tougher ones that additional take a look at efficiency. In the meantime, firms are investing more and more giant quantities of computational assets to coach laptop imaginative and prescient techniques at a quicker fee than ever earlier than. In the meantime, applied sciences to be used in deployed techniques — like object-detection frameworks for evaluation of nonetheless frames from movies — are maturing quickly, indicating additional AI deployment.Pure Language Processing (NLP) outruns its analysis metrics: Fast progress in NLP has yielded AI techniques with considerably improved language capabilities which have began to have a significant financial impression on the world. Google and Microsoft have each deployed the BERT language mannequin into their engines like google, whereas different giant language fashions have been developed by firms starting from Microsoft to OpenAI. Progress in NLP has been so swift that technical advances have began to outpace the benchmarks to check for them. This may be seen within the fast emergence of techniques that receive human-level efficiency on Superglue, an NLP analysis suite developed in response to earlier NLP progress overshooting the capabilities being assessed by GLUE.New analyses on reasoning: Most measures of technical issues present for every time level the efficiency of one of the best system at the moment on a set benchmark. New analyses developed for the AI Index provide metrics that permit for an evolving benchmark, and for the attribution of particular person techniques of credit score for a share of the general efficiency of a bunch of techniques over time. These are utilized to 2 symbolic reasoning issues, Automated Theorem Proving and Falsifiability of Boolean formulation.Machine studying is altering the sport in healthcare and biology: The panorama of the healthcare and biology industries has advanced considerably with the adoption of machine studying. DeepMind’s AlphaFold utilized deep studying strategies to make a big breakthrough within the decades-long organic problem of protein folding. Scientists use ML fashions to be taught representations of chemical molecules for simpler chemical synthesis planning. PostEra, an AI startup, used ML-based strategies to speed up COVID-related drug discovery through the pandemic.3. THE ECONOMY.“Medication, Most cancers, Molecular, Drug Discovery” obtained the best quantity of personal AI funding in 2020, with greater than USD 13.8 billion, 4.5 occasions greater than 2019.Brazil, India, Canada, Singapore, and South Africa are the nations with the very best progress in AI hiring from 2016 to 2020. Regardless of the COVID-19 pandemic, AI hiring continued to develop throughout pattern nations in 2020.Extra non-public funding in AI is being funneled into fewer startups. Regardless of the pandemic, 2020 noticed a 9.3% enhance within the quantity of personal AI funding from 2019 — a better proportion enhance than from 2018 to 2019 (5.7%), although the variety of newly based firms decreased for the third yr in a row.Regardless of rising calls to deal with moral considerations related to utilizing AI, efforts to deal with these considerations within the business are restricted, in line with a McKinsey survey. For instance, points resembling fairness and equity in AI proceed to obtain comparatively little consideration from firms. Furthermore, fewer firms in 2020 view private or particular person privateness dangers as related, in contrast inside 2019, and there was no change within the proportion of respondents whose firms are taking steps to mitigate these explicit dangers.Regardless of the financial downturn brought on by the pandemic, half of the respondents in a McKinsey survey stated that the coronavirus had no impact on their funding in AI, whereas 27% really reported rising their funding. Lower than a fourth of companies decreased their funding in AI.America recorded a lower in its share of AI job postings from 2019 to 2020 — the primary drop in six years. The overall variety of AI jobs posted in america additionally decreased by 8.2% from 2019 to 2020, from 325,724 in 2019 to 300,999 jobs in 2020.4: AI EDUCATION :• An AI Index survey carried out in 2020 means that the world’s high universities have elevated their funding in AI schooling over the previous 4 years. The variety of programs that train college students the talents mandatory to construct or deploy a sensible AI mannequin on the undergraduate and graduate ranges has elevated by 102.9% and 41.7%, respectively, within the final 4 educational years.• Extra AI Ph.D. graduates in North America selected to work within the business prior to now 10 years, whereas fewer opted for jobs in academia, in line with an annual survey from the Computing Analysis Affiliation (CRA). The share of latest AI PhDs who selected business jobs elevated by 48% prior to now decade, from 44.4% in 2010 to 65.7% in 2019. Against this, the share of latest AI PhDs coming into academia dropped by 44%, from 42.1% in 2010 to 23.7% in 2019.• Within the final 10 years, AI-related PhDs have gone from 14.2% of the entire of CS PhDs granted in america, to round 23% as of 2019, in line with the CRA survey. On the identical time, different beforehand in style CS PhDs have declined in reputation, together with networking, software program engineering, and programming languages. Compilers all noticed a discount in PhDs granted relative to 2010, whereas AI and Robotics/Imaginative and prescient specializations noticed a considerable enhance.• After a two-year enhance, the variety of AI college departing from universities to business jobs in North America dropped from 42 in 2018 to 33 in 2019 (28 of those are tenured college and 5 are untendered). Carnegie Mellon College had the most important variety of AI college departures between 2004 and 2019 (16), adopted by the Georgia Institute of Expertise (14) and the College of Washington (12).The proportion of worldwide college students amongst new AI PhDs in North America continued to rise in 2019, to 64.3% — a 4.3% enhance from 2018. Amongst international graduates, 81.8% stayed in america and eight.6% have taken jobs outdoors america.Within the European Union, the overwhelming majority of specialised AI educational choices are taught on the grasp’s stage; robotics and automation is by far essentially the most often taught course within the specialised bachelor’s and grasp’s packages, whereas machine studying (ML) dominates within the specialised brief programs.5: ETHICAL CHALLENGES OF AI APPLICATIONSThe variety of papers with ethics-related key phrases in titles submitted to AI conferences has grown since 2015, although the common variety of paper titles matching ethics-related key phrases at main AI conferences stays low through the years.The 5 information matters that received essentially the most consideration in 2020 associated to the moral use of AI have been the discharge of the European Fee’s white paper on AI, Google’s dismissal of ethics researcher Tim nit Gebru, the AI ethics committee fashioned by the United Nations, the Vatican’s AI ethics plan, and IBM’s exiting the facial-recognition companies.6: DIVERSITY IN AI• The odds of feminine AI Ph.D. graduates and tenure-track laptop science (CS) college have remained low for greater than a decade. Feminine graduates of AI Ph.D. packages in North America have accounted for lower than 18% of all Ph.D. graduates on common, in line with an annual survey from the Computing Analysis Affiliation (CRA). An AI Index survey means that feminine college make up simply 16% of all tenure-track CS college at a number of universities world wide.• The CRA survey means that in 2019, amongst new U.S. resident AI Ph.D. graduates, 45% have been white, whereas 22.4% have been Asian, 3.2% have been Hispanic, and a pair of.4% have been African American.• The proportion of white (non-Hispanic) new computing PhDs has modified little over the past 10 years, accounting for 62.7% on common. The share of Black or African American (non-Hispanic) and Hispanic computing PhDs in the identical interval is considerably decrease, with a median of three.1% and three.3%, respectively.The participation of Black in AI workshops, that are co-located with the Convention on Neural Data Processing Techniques (NeurIPS), has grown considerably lately. The variety of attendees and papers submitted in 2019 is 2.6 occasions greater than in 2017, whereas the variety of accepted papers is 2.1 occasions greater.In a membership survey by Queer in AI in 2020, virtually half the respondents stated they view the dearth of inclusiveness within the discipline as an impediment they’ve confronted in turning into a practitioner within the AI/ML discipline. Greater than 40% of members surveyed stated they’ve skilled discrimination or harassment at work or faculty.7: AI POLICY AND NATIONAL STRATEGIES:Since Canada revealed the world’s first nationwide AI technique in 2017, greater than 30 different nations and areas have revealed related paperwork as of December 2020.The launch of the International Partnership on AI (GPAI) and Organisation for Financial Co-operation and Growth (OECD) AI Coverage Observatory and Community of Specialists on AI in 2020 promoted intergovernmental efforts to work collectively to help the event of AI for all.In america, the 116th Congress was essentially the most AI-focused congressional session in historical past. The variety of mentions of AI by this Congress in laws, committee reviews, and Congressional Analysis Service (CRS) reviews is greater than triple that of the one hundred and fifteenth Congress.Reference:[1] Daniel Zhang, Saurabh Mishra, Erik Brynjolfsson, John Etchemendy, Deep Ganguli, Barbara Grosz, Terah Lyons, James Manyika, Juan Carlos Niebles, Michael Sellitto, Yoav Shoham, Jack Clark, and Raymond Perrault, “The AI Index 2021 Annual Report,” AI Index Steering Committee, Human-Centered AI Institute, Stanford College, Stanford, CA, March 2021.

Article Tags:
Article Categories:
Artificial · Intelligence

Leave a Reply

Your email address will not be published. Required fields are marked *