Getting AI Right
Introductory Notes on AI & Society
Bibliographic Data
| ID | 9684827 |
|---|---|
| Authors | James Manyika (corresponding author) |
| Year | 2022 |
| Volume | 151 |
| Issue | 2 |
| Pages | 5-27 |
| Publication date | 2022-05-01 |
| Peer Reviewed | Yes |
| Open Access | Yes |
| Type | ARTICLE |
| Venue | Daedalus (JOURNAL) |
| Journal identifiers | ISSN: 0011-5266 • E-ISSN: 1548-6192 |
| Publisher | MIT Press (PUBLISHER • US) |
| DOI | 10.1162/daed_e_01897 |
| OpenAlex | W4226016139 |
| Language | EN |
| Citations received | 5 |
| References cited | 5 |
This dialogue is from an early scene in the 2014 film Ex Machina, in which Nathan has invited Caleb to determine whether Nathan has succeeded in creating artificial intelligence.1 The achievement of powerful artificial general intelligence has long held a grip on our imagination not only for its exciting as well as worrisome possibilities, but also for its suggestion of a new, uncharted era for humanity. In opening his 2021 BBC Reith Lectures, titled “Living with Artificial Intelligence,” Stuart Russell states that “the eventual emergence of general-purpose artificial intelligence [will be] the biggest event in human history.”2Over the last decade, a rapid succession of impressive results has brought wider public attention to the possibilities of powerful artificial intelligence. In machine vision, researchers demonstrated systems that could recognize objects as well as, if not better than, humans in some situations. Then came the games. Complex games of strategy have long been associated with superior intelligence, and so when AI systems beat the best human players at chess, Atari games, Go, shogi, StarCraft, and Dota, the world took notice. It was not just that Als beat humans (although that was astounding when it first happened), but the escalating progression of how they did it: initially by learning from expert human play, then from self-play, then by teaching themselves the principles of the games from the ground up, eventually yielding single systems that could learn, play, and win at several structurally different games, hinting at the possibility of generally intelligent systems.3Speech recognition and natural language processing have also seen rapid and headline-grabbing advances. Most impressive has been the emergence recently of large language models capable of generating human-like outputs. Progress in language is of particular significance given the role language has always played in human notions of intelligence, reasoning, and understanding. While the advances mentioned thus far may seem abstract, those in driverless cars and robots have been more tangible given their embodied and often biomorphic forms. Demonstrations of such embodied systems exhibiting increasingly complex and autonomous behaviors in our physical world have captured public attention.Also in the headlines have been results in various branches of science in which AI and its related techniques have been used as tools to advance research from materials and environmental sciences to high energy physics and astronomy.4 A few highlights, such as the spectacular results on the fifty-year-old protein-folding problem by AlphaFold, suggest the possibility that AI could soon help tackle science's hardest problems, such as in health and the life sciences.5While the headlines tend to feature results and demonstrations of a future to come, AI and its associated technologies are already here and pervade our daily lives more than many realize. Examples include recommendation systems, search, language translators - now covering more than one hundred languages - facial recognition, speech to text (and back), digital assistants, chatbots for customer service, fraud detection, decision support systems, energy management systems, and tools for scientific research, to name a few. In all these examples and others, AI-related techniques have become components of other software and hardware systems as methods for learning from and incorporating messy real-world inputs into inferences, predictions, and, in some cases, actions. As director of the Future of Humanity Institute at the University of Oxford, Nick Bostrom noted back in 2006, “A lot of cutting-edge AI has filtered into general applications, often without being called AI because once something becomes useful enough and common enough it's not labeled AI anymore.”6As the scope, use, and usefulness of these systems have grown for individual users, researchers in various fields, companies and other types of organizations, and governments, so too have concerns when the systems have not worked well (such as bias in facial recognition systems), or have been misused (as in deepfakes), or have resulted in harms to some (in predicting crime, for example), or have been associated with accidents (such as fatalities from self-driving cars).7Dædalus last devoted a volume to the topic of artificial intelligence in 1988, with contributions from several of the founders of the field, among others. Much of that issue was concerned with questions of whether research in AI was making progress, of whether AI was at a turning point, and of its foundations, mathematical, technical, and philosophical-with much disagreement. However, in that volume there was also a recognition, or perhaps a rediscovery, of an alternative path toward AI - the connectionist learning approach and the notion of neural nets-and a burgeoning optimism for this approach's potential. Since the 1960s, the learning approach had been relegated to the fringes in favor of the symbolic formalism for representing the world, our knowledge of it, and how machines can reason about it. Yet no essay captured some of the mood at the time better than Hilary Putnam's “Much Ado About Not Very Much.” Putnam questioned the Dædalus issue itself: “Why a whole issue of Dædalus? Why don't we wait until AI achieves something and then have an issue?” He concluded:This volume of Dædalus is indeed the first since 1988 to be devoted to artificial intelligence. This volume does not rehash the same debates; much else has happened since, mostly as a result of the success of the machine learning approach that was being rediscovered and reimagined, as discussed in the 1988 volume. This issue aims to capture where we are in and how its The and concerns are by with the and that in AI in in as an at the University of a research to and a neural on to research on AI and at the have been with researchers in and AI systems, on the progress, and with in and with its and for of the in this volume from AI and at the of many of to at the of on The volume is into of the are on the other on its with various of In to the in their and the a of on the possibilities, and concerns for to the for to these it may be useful to we by artificial intelligence. The headlines and of AI and its associated technologies have to some and about as This has not been by the researchers in science and the and not only machine but and of all with This could the now associated with but it could also be an of the success of the of AI and its related techniques and their and are but it has not always been In the now to as the AI which in AI did not to there was a to of we now AI with types of are given for The first are those that suggest that it is the to intelligent can artificial intelligence human in such include speech recognition, the to problems, and from of this are by some to be in their as to as intelligence and in the for success they for the of AI on this The of to be of and an intelligent or its or of also the of which could be given to the or types of are this volume of its the much to the of some in the field, has to be and called a It is on with for intelligence, those on and neural and various other and as well as their in and, in the of embodied intelligence, systems that can and questions the in this are we in and does AI for much about AI is about of intelligent machines all the back to among and have been about AI for a long that may have the The of machine intelligence to of the machine in the and to the of several of his in the the of artificial intelligence as we it and the of the is generally to the now of The was the result of a for a by and be to how to machines and of now for and their contributions to this a of Artificial and and in different but and Stuart Russell the and in its of as well as the AI The AI has been since the with headline-grabbing in rapid succession the last or a that in the of his essay as a not only for the of AI but also its in a of of as well as of scientific This is best by the approach to artificial intelligence learning from and by the success of neural and with methods from as for machines to may be useful In the there of how to machine intelligence. was to to a and symbolic of the world and our knowledge of it and, from systems that could reason about the world, thus exhibiting intelligence to the This was by and with and others. associated with it was the approach that intelligence was a problem of a of possibilities for The was by the than the and to intelligence by In as the connectionist called in by the of in the the this approach was associated with While there was about the first came to and did so for with some expert only did this approach from by its and it came with the of a long by and among to and to and knowledge and It was only in the that to in the vision, the of and others. The of these and the associated are discussed in and Stuart 1988 Dædalus essay a the Artificial at a Since the approach to intelligence on the of and and has to the in his essay Not It It Artificial of the of and and with the of machines that can not only learn, but and they learn, and the of this in the systems now being and those to The success of the machine learning approach has from the in the of to the to the in the of the and other and In research, the has been the result of scientific and and for in and in has been the of the software and hardware better to the in and neural and other machine learning as into in of In their for and in machine and the of such as that could be used for In their and of Intelligence,” and different and in natural language the emergence of large language models of of and that and learning on of The models are impressive in their to natural language for which they have not been and human-like not only in natural but also software and as and in have to to these large language models as models in that once they are they are to a of and their these large language models are early in their and have many and that are in this volume and by some of their from the in systems, advances in the as well as in their that to in the physical the in the used thus far and of robots that on from In as in AI more there has always been a as to whether to or from how humans and other intelligent AI and have how and AI from and so far more in one than the as and have the success of the to there are many and as well as in It is useful to on one such as when AI does not as or or or that can to or when it on or about the world, or when it has such as of all of which can to a of public have captured the attention of the wider public and as well as among there is an on AI and In there has been a of to principles and to as well as and such as the on that to best has been the of with to and - in the and AI in and as has been well in This is an in its but also with to the of the AI and, in its with more the other there are and associated with the that AI is not capable of if could to more more or more general In their learning and took of where learning and its such as the with In the of natural language and the in and the of large language and have the notion that large language models or In of in a and the in systems, the as how to reason about other their systems, and well as in and when the include humans and and a useful of the in there is a among many that we not have for the of AI systems, as they become more capable and the of AI and its related techniques are to be powerful tools for research in as examples in this volume and examples in which AI not only help results but also by and become some have AI to science and and to and for the possibility that more powerful AI could to in as well as in some of and has long been a for many at the of AI research to more capable the of of the of more general that to the possibility of more capable AI reasoning, of and and of these and other that could to more capable systems the of whether by the of and and more and and or whether different are in such as or or on and to name a few. whether and of be the AI is but many the with of and learning have to their about the of the is associated with the of whether artificial general intelligence can be and if how and Artificial general intelligence is in to is called that AI and for and such as The of on the other aims for more powerful AI - at as powerful as is generally to problem or and, in some the to and as well as and its and the of and when be is a for that its achievement have and as is often in and such as A and The to Ex and it is or there is among many at the of AI research that we for the possibility of powerful with to and and with its and use, and the possibility that of could and that we these into how we approach the of of the research and and in AI is of the AI and in its the of This is given the for useful and and the for in of the However, a few have the of their the of these are and of which has demonstrated results of a long from the discussed of AI and is on and the future of This is not In in the of the about AI and and concerns about their on a on and the was that such technologies for and and “the that but not Most of this those have been have and that more are than are that it is the and the and the of the In their essay AI and these for and in the of to the with to and and as well as the that are in In The of Artificial Intelligence,” how the of human in the of AI the of AI that than human He that the in this and for on the for and a that the that more be than too much from of the and does not far enough into the future and at AI be capable The for AI could from of that in the is and and to are and and until has mostly physical and but that AI be on more and on and, if early examples are are not of the In other are now in the world machines that that and that their to these is to a of they can be with the to which the human has been This was and in that this time could be different that in which by other humans for their when machines may be capable of these as well as or better than The other is that AI so much and all without the for human and the of be to for when that the that once the first time since his be with his his to his from how to the which science and have for to and and However, researchers that we are not to a future in which the of and that until there are other and that be in the now and in the such as and other and how humans increasingly capable that and and in this are not the only of the by Russell a of the from artificial general intelligence, once a of or we to general-purpose the for companies and, for the and as well as from AI and its related technologies are more than to and by companies and in the and of the many the is it is generally that is a in as by its in AI research, and as in several have for companies and given the of such technologies as discussed by and the may in the of to AI and (such as whether they are companies or as and have have the to to in The role of AI in intelligence, systems, autonomous and other of increasingly In the on Artificial a of and future that AI technologies to and for the of and on among of to to and the same Russell and concerns autonomous and of from the of for AI from other such as on the of However, there are AI and for has been by a in of as AI technologies in their and have and tend to those in their whether they are companies or This and for companies and on one and the of on the Not in this volume are the for companies (as well as other types of and not at the of but that could from its This is of particular significance given that many have the for AI and its related technologies to with other and to many and future and The has given a of the human when at the of a such as a not or are to it with of the the of AI has grown to not only and but also those in health public and it has in many and decision and and the same in some cases, the of has to of bias and often the result of bias in the and the systems which such are examples from facial recognition, and to in in an of Artificial when there is an of and systems the same and for as In of Artificial from and the of ground and when for and are at with of and the of In of these as well as the possibilities of a and how we the of their intelligent is to how and in their to and be to AI to increasingly powerful In the and of when well and the that when are the for how AI governments, the to AI to the and of public is also to In essay AI for a public AI on public examples from different to those related to that are more in the public than they are for by how and in which can the and of the of from his Dædalus artificial intelligence in as a of at the human of the success of the themselves in various of human this one can various of questions about the AI as it becomes more does it to be more human when many of the we have these of increasingly can be or in the better or better by much of being human the of not how it or on our to it or it when this our human of could be intelligence, and so the possibility of other of that may or we our systems and our at the of our at our notions of and and to they are (and have and they to be in the of the emergence of large language models and their to human-like a of as it, to questions about in an era of increasingly capable may have at at the of his 1988 that has not of our but it has with of and imagination that we have only just to how humans could to and machines when machines can human-like this with examples by large language The possibilities that we humans may have to our to and other the of his essay a of with from large language He whether we are or the when it to notions such as intelligence or in to these for in of the field, first in it of the of the of artificial intelligence that time how to a but was a of to not As to machines are or not when they to be one could whether they are is different from humans in other than how it is being In as a in about the of these also how of the human have been with language in different and the possibility of a era in which language is from for the of technologies have played a and role in our of and the In how they are and and by that technologies AI have the of particular of it is to be human and to progress, often at the of other of being human and of and those associated with and and with and the for among those in has other technologies with the to AI can also of already In of AI as to and the of machine learning to physical with not and and the of that can result when AI is used as a of in AI not only the on associated with its it also to already in have how and can not only but also and as well as In the to bias or for may also the for to in a it by it is being by Putnam's of that has a to that has to the role of and and the to may on and in the first that have to attention in the of and In a of for Artificial Intelligence,” on of the of AI and by the role that systems He and of to suggest that of may now to with the of AI as a and that could to some as AI becomes more a becomes how to AI with humans with to and and The of AI and is as as the it, as Russell researchers that concerns about these of are given the of we are a long from but that research to to the issue and it into how we more powerful AI Russell an approach to and human that on in and human and of learning as a for machines to human has discussed the of possibilities as to we by with with possibility its in as in there are with to the of human that AI that with human is one but it does not the to determine those human be in the first He the to to and its by AI as well as by and often and as which to at the of and some of it that the by more capable AI highlights, and with the to be of not how AI may be to and that are no just the of since we are autonomous systems that may have to such the of we have how is it This has been about from or Go, to a a creating making or scientific or a In the it may be useful also to to an toward for one AI with the humans best at the of in which the humans better than AI - as they for the is well by these humans these also with other of such as the in or to that or a as AI becomes more is where the more we for humans to with machine when AI is A be with to how the of AI become with those of and as mentioned eventually to AI systems become in this the of a era for and for of the discussed perhaps the is with to we to AI for and we AI to be capable of in to It seem that in such with how we and the are not but that are to all this a be to and of the back in This given the and and the that has been and not to the usefulness of AI to its and the possibilities, some already for to the is a an AI a few on a to which the Stuart a the or how but have to it as aims to the possibilities we the we the we the we the and we and the we all in to the in a In other it is a of we to if AI is to be a to in this volume of Dædalus many of the we from these and other and a one can a long can be as The first is related to the of AI powerful and capable enough to the exciting possibilities for but also and without or individual or and to public where are A of concerns and where it can the contributions to as in health and the life and in the sciences and in scientific to for all The all is given the that without attention to it, the of the AI and its could to a few organizations, and those in its and The of on the use, and of This is given the and and the for in AI that has been companies and as a Not to AI could to and and many more and among the various A of concerns how we our systems and and the of how to be human in an of increasingly powerful of this volume their on we if AI is to be a for humanity. While such as our and with AI and as AI becomes more the on not to the is this is on and these it that the of human and and other are on and the of and not enough on the other that are also for AI to be a to given the can to the for the to this Dædalus volume on AI and to on AI a of a as as there are without many more and that are for that to the of Oxford, where have been a the of this Dædalus volume. also to at the the AI and the of AI at as well as on the of and on and for our many as well as our that the of this volume. for the with the in this volume and with others, and for and on this from and Stuart but they not be held for or volume could not have without the of the of of and of brought as and and from the to the of this and and and expert for all the in this volume
Political science · Ethics and Social Impacts of AI · Philosophy
| Unique citing works | 5 |
|---|---|
| Citations per year | 1,67 |
| Citation span | 2023 - 2025 (3) |
| Citation velocity | recent |
| Highly cited | No |
| Citation types | Neutral: 5 |