Hubert Dreyfus was a critic ofartificial intelligence research. In a series of papers and books, includingAlchemy and AI(1965),What Computers Can't Do (1972;1979;1992) andMind over Machine(1986), he presented a pessimistic assessment of AI's progress and a critique of the philosophical foundations of the field. Dreyfus' objections are discussed in most introductions to thephilosophy of artificial intelligence, includingRussell & Norvig (2021), a standard AI textbook, and inFearn (2007), a survey of contemporary philosophy.[1]
Dreyfus argued that human intelligence and expertise depend primarily on unconscious processes rather than conscioussymbolic manipulation, and that these unconscious skills can never be fully captured in formal rules. His critique was based on the insights of moderncontinental philosophers such asMerleau-Ponty andHeidegger, and was directed at the first wave of AI research which used high levelformal symbols to represent reality and tried to reduce intelligence to symbol manipulation.
When Dreyfus' ideas were first introduced in the mid-1960s, they were met with ridicule and outright hostility.[2][3] By the 1980s, however, many of his perspectives were rediscovered by researchers working inrobotics and the new field ofconnectionism—approaches now called "sub-symbolic" because they eschew early AI research's emphasis on high level symbols. In the 21st century,statistics-based approaches tomachine learning simulate the way that the brain uses unconscious process to perceive, notice anomalies and make quick judgements. These techniques are highly successful and are currently widely used in both industry and academia. Historian and AI researcherDaniel Crevier writes: "time has proven the accuracy and perceptiveness of some of Dreyfus's comments."[4] Dreyfus said in 2007, "I figure I won and it's over—they've given up."[5]
InAlchemy and AI(1965) andWhat Computers Can't Do(1972),Dreyfus summarized thehistory of artificial intelligence and ridiculed the unbridled optimism that permeated the field. For example,Herbert A. Simon, following the success of his programGeneral Problem Solver(1957), predicted that by 1967:[6]
The press reported these predictions in glowing reports of the imminent arrival of machine intelligence.
Dreyfus felt that this optimism was unwarranted and based on false assumptions about the nature of human intelligence.Pamela McCorduck explains Dreyfus' position:
A great misunderstanding accounts for public confusion about thinking machines, a misunderstanding perpetrated by the unrealistic claims researchers in AI have been making, claims that thinking machines are already here, or at any rate, just around the corner.[7]
These predictions were based on the success of an "information processing" model of the mind, articulated by Newell and Simon in theirphysical symbol systems hypothesis, and later expanded into a philosophical position known ascomputationalism by philosophers such asJerry Fodor andHilary Putnam.[8] Believing that they had successfully simulated the essential process of human thought with simple programs, it seemed a short step to producing fully intelligent machines. However, Dreyfus argued that philosophy, especially20th-century philosophy, had discovered serious problems with this information processing viewpoint. The mind, according to modern philosophy, is nothing like a digital computer.[7]
InAlchemy and AI andWhat Computers Can't Do, Dreyfus identified four philosophical assumptions that supported the faith of early AI researchers that human intelligence depended on the manipulation of symbols.[9] "In each case," Dreyfus writes, "the assumption is taken by workers in [AI] as an axiom, guaranteeing results, whereas it is, in fact, one hypothesis among others, to be tested by the success of such work."[10]
In the early days of research intoneurology, scientists realized thatneurons fire in all-or-nothing pulses. Several researchers, such asWalter Pitts andWarren McCulloch, argued that neurons functioned similar to the wayBoolean logic gates operate, and so could be imitated by electronic circuitry at the level of the neuron.[11]When digital computers became widely used in the early 50s, this argument was extended to suggest that the brain was a vastphysical symbol system, manipulating the binary symbols of zero and one. Dreyfus was able to refute the biological assumption by citing research inneurology that suggested that the action and timing of neuron firing had analog components.[12] But Daniel Crevier observes that "few still held that belief in the early 1970s, and nobody argued against Dreyfus" about the biological assumption.[13]
He refuted this assumption by showing that much of what we "know" about the world consists of complexattitudes ortendencies that make us lean towards one interpretation over another. He argued that, even when we use explicit symbols, we are using them against an unconscious background ofcommonsense knowledge and that without this background our symbols cease to mean anything. This background, in Dreyfus' view, was not implemented in individual brains as explicit individual symbols with explicit individual meanings.
This concerns the philosophical issue ofepistemology, or the study ofknowledge. Even if we agree that the psychological assumption is false, AI researchers could still argue (as AI founderJohn McCarthy has) that it is possible for a symbol processing machine to represent all knowledge, regardless of whether human beings represent knowledge the same way. Dreyfus argued that there is no justification for this assumption, since so much of human knowledge is not symbolic.
Dreyfus also identified a subtler assumption about the world. AI researchers (and futurists and science fiction writers) often assume that there is no limit to formal, scientific knowledge, because they assume that any phenomenon in the universe can be described by symbols or scientific theories. This assumes that everything thatexists can be understood as objects, properties of objects, classes of objects, relations of objects, and so on: precisely those things that can be described by logic, language and mathematics. The study of being or existence is calledontology, and soDreyfus calls this the ontological assumption. If this is false, then it raises doubts about what we can ultimately know and what intelligent machines will ultimately be able to help us to do.
InMind Over Machine(1986), written during the heyday ofexpert systems, Dreyfus analyzed the difference between human expertise and the programs that claimed to capture it. This expanded on ideas fromWhat Computers Can't Do, where he had made a similar argument criticizing the "cognitive simulation" school of AI research practiced byAllen Newell andHerbert A. Simon in the 1960s.
Dreyfus argued that human problem solving and expertise depend on our background sense of the context, of what is important and interesting given the situation, rather than on the process of searching through combinations of possibilities to find what we need. Dreyfus would describe it in 1986 as the difference between "knowing-that" and "knowing-how", based onHeidegger's distinction ofpresent-at-hand andready-to-hand.[14]
Knowing-that is our conscious, step-by-step problem solving abilities. We use these skills when we encounter a difficult problem that requires us to stop, step back and search through ideas one at time. At moments like this, the ideas become very precise and simple: they become context free symbols, which we manipulate using logic and language. These are the skills thatNewell andSimon had demonstrated with both psychological experiments and computer programs. Dreyfus agreed that their programs adequately imitated the skills he calls "knowing-that."
Knowing-how, on the other hand, is the way we deal with things normally. We take actions without using conscious symbolic reasoning at all, as when we recognize a face, drive ourselves to work or find the right thing to say. We seem to simply jump to the appropriate response, without considering any alternatives. This is the essence of expertise, Dreyfus argued: when our intuitions have been trained to the point that we forget the rules and simply "size up the situation" and react.
The human sense of the situation, according to Dreyfus, is based on our goals, our bodies and our culture—all of our unconscious intuitions, attitudes and knowledge about the world. This “context” or "background" (related toHeidegger'sDasein) is a form of knowledge that is not stored in our brains symbolically, but intuitively in some way. It affects what we notice and what we don't notice, what we expect and what possibilities we don't consider: we discriminate between what is essential and inessential. The things that are inessential are relegated to our "fringe consciousness" (borrowing a phrase fromWilliam James): the millions of things we're aware of, but we're not really thinking about right now.
Dreyfus does not believe that AI programs, as they were implemented in the 70s and 80s, could capture this "background" or do the kind of fast problem solving that it allows. He argued that our unconscious knowledge couldnever be captured symbolically. If AI could not find a way to address these issues, then it was doomed to failure, an exercise in "tree climbing with one's eyes on the moon."[15]
Dreyfus began to formulate his critique in the early 1960s while he was a professor atMIT, then a hotbed of artificial intelligence research. His first publication on the subject is a half-page objection to a talk given byHerbert A. Simon in the spring of 1961.[16] Dreyfus was especially bothered, as a philosopher, that AI researchers seemed to believe they were on the verge of solving many long standing philosophical problems within a few years, using computers.
In 1965, Dreyfus was hired (with his brotherStuart Dreyfus' help) byPaul Armer to spend the summer atRAND Corporation's Santa Monica facility, where he would write "Alchemy and AI", the first salvo of his attack. Armer had thought he was hiring an impartial critic and was surprised when Dreyfus produced a scathing paper intended to demolish the foundations of the field. (Armer stated he was unaware of Dreyfus' previous publication.) Armer delayed publishing it, but ultimately realized that "just because it came to a conclusion you didn't like was no reason not to publish it."[17] It finally came out as RAND Memo and soon became a best seller.[18]
The paper flatly ridiculed AI research, comparing it toalchemy: a misguided attempt to change metals to gold based on a theoretical foundation that was no more than mythology and wishful thinking.[19] It ridiculed the grandiose predictions of leading AI researchers, predicting that there were limits beyond which AI would not progress and intimating that those limits would be reached soon.[20]
The paper "caused an uproar", according to Pamela McCorduck.[21] The AI community's response was derisive and personal.Seymour Papert dismissed one third of the paper as "gossip" and claimed that every quotation was deliberately taken out of context.[22]Herbert A. Simon accused Dreyfus of playing "politics" so that he could attach the prestigious RAND name to his ideas. Simon said, "what I resent about this was the RAND name attached to that garbage".[23]
Dreyfus, who taught atMIT, remembers that his colleagues working in AI "dared not be seen having lunch with me."[24]Joseph Weizenbaum, the author ofELIZA, felt his colleagues' treatment ofDreyfus was unprofessional and childish. Although he was an outspoken critic of Dreyfus' positions, he recalls "I became the only member of the AI community to be seen eating lunch with Dreyfus. And I deliberately made it plain that theirs was not the way to treat a human being."[25]
The paper was the subject of a short inThe New Yorker magazine on June 11, 1966. The piece mentioned Dreyfus' contention that, while computers may be able to play checkers, no computer could yet play a decent game of chess. It reported with wry humor (as Dreyfus had) about the victory of a ten-year-old over the leading chess program, with "even more than its usual smugness."[20]
In hope of restoring AI's reputation,Seymour Papert arranged a chess match between Dreyfus andRichard Greenblatt'sMac Hack program. Dreyfus lost, much to Papert's satisfaction.[26] AnAssociation for Computing Machinery bulletin[27] used the headline:
Dreyfus complained in print that he hadn't said a computer willnever play chess, to whichHerbert A. Simon replied: "You should recognize that some of those who are bitten by your sharp-toothed prose are likely, in their human weakness, to bite back ... may I be so bold as to suggest that you could well begin the cooling---a recovery of your sense of humor being a good first step."[29]
By the early 1990s several of Dreyfus' radical opinions had become mainstream.
Failed predictions. As Dreyfus had foreseen, the grandiose predictions of early AI researchers failed to come true. Fully intelligent machines (now known as "strong AI") did not appear in the mid-1970s as predicted.HAL 9000 (whose capabilities for natural language, perception and problem solving were based on the advice and opinions ofMarvin Minsky) did not appear in the year 2001. "AI researchers", writes Nicolas Fearn, "clearly have some explaining to do."[30] Today researchers are far more reluctant to make the kind of predictions that were made in the early days. (Although some futurists, such asRay Kurzweil, are still given to the same kind of optimism.)
The biological assumption, although common in the forties and early fifties, was no longer assumed by most AI researchers by the time Dreyfus publishedWhat Computers Can't Do.[13] Although many still argue that it is essential to reverse-engineer the brain by simulating the action of neurons (such asRay Kurzweil[31] orJeff Hawkins[32]), they don't assume that neurons are essentially digital, but rather that the action of analog neurons can be simulated by digital machines to a reasonable level of accuracy.[31] (Alan Turing had made this same observation as early as 1950.)[33]
The psychological assumption andunconscious skills. Many AI researchers have come to agree that human reasoning does not consist primarily of high-level symbol manipulation. In fact, since Dreyfus first published his critiques in the 60s, AI research in general has moved away fromhigh level symbol manipulation, towards new models that are intended to capture more of ourunconscious reasoning.Daniel Crevier writes that by 1993, unlike 1965, AI researchers "no longer made the psychological assumption",[13] and had continued forward without it.
In the 1980s, these new "sub-symbolic" approaches included:
In the 1990s and the early decades of the 21st century,statistics-based approaches to machine learning used techniques related to economics and statistics to allow machines to "guess" – to make inexact, probabilistic decisions and predictions based on experience and learning. These programs simulate the way our unconscious instincts are able to perceive, notice anomalies and make quick judgements, similar to what Dreyfus called "sizing up the situation and reacting", but here the "situation" consists of vast amounts of numerical data. These techniques are highly successful and are currently widely used in both industry and academia.
This research has gone forward without any direct connection to Dreyfus' work.[36]
Knowing-how and knowing-that. Research in psychology and economics has been able to show that Dreyfus' (and Heidegger's) speculation about the nature of human problem solving was essentially correct.Daniel Kahnemann andAmos Tversky collected a vast amount of hard evidence that human beings use two very different methods to solve problems, which they named "system 1" and "system 2". System one, also known as theadaptive unconscious, is fast, intuitive and unconscious. System 2 is slow, logical and deliberate. Their research was collected in the bookThinking, Fast and Slow,[37] and inspiredMalcolm Gladwell's popular bookBlink.[38] As with AI, this research was entirely independent of both Dreyfus and Heidegger.[36]
Although clearly AI research has come to agree with Dreyfus, McCorduck claimed that "my impression is that this progress has taken place piecemeal and in response to tough given problems, and owes nothing to Dreyfus."[36]
The AI community, with a few exceptions, chose not to respond to Dreyfus directly. "He's too silly to take seriously" a researcher told Pamela McCorduck.[29]Marvin Minsky said of Dreyfus (and the other critiques coming fromphilosophy) that "they misunderstand, and should be ignored."[39] When Dreyfus expandedAlchemy and AI to book length and published it asWhat Computers Can't Do in 1972, no one from the AI community chose to respond (with the exception of a few critical reviews). McCorduck asks "If Dreyfus is so wrong-headed, why haven't the artificial intelligence people made more effort to contradict him?"[29]
Part of the problem was thekind of philosophy that Dreyfus used in his critique. Dreyfus was an expert inmodern European philosophers (likeHeidegger andMerleau-Ponty).[40] AI researchers of the 1960s, by contrast, based their understanding of the human mind on engineering principles and efficient problem solving techniques related tomanagement science. On a fundamental level, they spoke a different language.Edward Feigenbaum complained, "What does he offer us?Phenomenology! That ball of fluff. That cotton candy!"[41] In 1965, there was simply too huge a gap between European philosophy andartificial intelligence, a gap that has since been filled bycognitive science,connectionism androbotics research. It would take many years before artificial intelligence researchers were able to address the issues that were important to continental philosophy, such assituatedness,embodiment,perception andgestalt.
Another problem was that he claimed (or seemed to claim) that AI wouldnever be able to capture the human ability to understand context, situation or purpose in the form of rules. But (asPeter Norvig andStuart Russell would later explain), an argument of this form cannot be won: just because one cannot imagine formal rules that govern human intelligence and expertise, this does not mean that no such rules exist. They quoteAlan Turing's answer to all arguments similar to Dreyfus's:
"we cannot so easily convince ourselves of the absence of complete laws of behaviour ... The only way we know of for finding such laws is scientific observation, and we certainly know of no circumstances under which we could say, 'We have searched enough. There are no such laws.'"[42]
Dreyfus did not anticipate that AI researchers would realize their mistake and begin to work towards new solutions, moving away from the symbolic methods that Dreyfus criticized. In 1965, he did not imagine that such programs would one day be created, so he claimed AI was impossible. In 1965, AI researchers did not imagine that such programs were necessary, so they claimed AI was almost complete. Both were wrong.
A more serious issue was the impression that Dreyfus' critique was incorrigibly hostile. McCorduck wrote, "His derisiveness has been so provoking that he has estranged anyone he might have enlightened. And that's a pity."[36] Daniel Crevier stated that "time has proven the accuracy and perceptiveness of some of Dreyfus's comments. Had he formulated them less aggressively, constructive actions they suggested might have been taken much earlier."[4]
In 2007, Dreyfus published a paper titled "Why Heideggerian AI Failed and How Fixing it Would Require Making It More Heideggerian", in which he reassesses the contemporary status of AI.[43] While acknowledging efforts at reorienting artificial intelligence research beyond "Good Old Fashioned AI" (GOFAI) towards "Heideggerian AI", he offers new criticisms on the shortcomings of existing programs.
In particular, Dreyfus considers Brooks's "behavior-based robots" andPhil Agre and David Chapman's "Pengi" as recent examples of Heideggerian AI. He rejects both programs as having neglected the dimension of learning new relevance in a worldly context. This has also been referred to as the "frame problem".
On the one hand, Dreyfus criticizesRodney Brooks's robots for "respond[ing] only to *fixed features* of the environment", and therefore are merely "converting stimulus input into reflex responses". It contrasts with the ability to comprehend and effectively learn the "world" as a context of practical significance, or "what-for", to human intelligence, as opposed to merely a totality of things. This echoes withHeidegger's comparison between the "world-poor" quality of animals versus human Dasein's capacity of "world-making" inThe Principle of Reason.[44]
On the other, Pengi also fails to meet Dreyfus's criterion of a genuinely Heideggerian AI. While Dreyfus praises Agre's understanding of "readiness-to-hand" as functions rather than entities, the program nevertheless involves "no skill [...] and no learning", but instead deterministically triggers responses that are evaluated based on set benchmark. The remark coincides with Dreyfus's commentary onDeep Blue's victory against Garry Kasparov, which noted that the machine is only responding to the "isolated domain" of chess, as opposed to the full range of possibilities of the "rest of the human life". While recent developments inartificial general intelligence (AGI) may have invalidated Dreyfus's objection, Fjelland (2020) argued that a disembodied AI could never handle "causal questions" that depend on a model of reality.[45]
Dreyfus then moves on to consider theMichael Wheeler's "embodied-embedded paradigm" as a refinement of Heideggerian AI. He accuses Wheeler of a "cognitivst misreading" for understanding Heidegger's insight to be the human capacity to use representational equipments external to the body. Instead, "being-in-the-world" is non-representationalist on the most basic level as a background to coping. The mind, therefore, does not "extended" into the background but exists within it. However, this objection has also been challenged by others as resting on an absolute dichotomy between reflective and pre-reflective modes of intentional behavior.
Finally, Dreyfus examinesWalter Freeman's program of Heideggerian neurodynamics, which sought to provide the material basis to phenomenological concepts in terms of a self-organizing neural "perception-action loop". In this model, relevant brain states are "attracted" to each other, while the brain switches from one equilibrium to another. The resulting dynamic system characterizes the "intentional arc" that affords contextual significance to actions. One complication Dreyfus does not address, however, is that this "neurodynamics of intention" is not distinctive of human intelligence, but as e.g. Freeman (2015) has argued, evolved from "[...] Ordovician period as a tool to prowl first olfactory environments, then environments of other modalities."[46]
He concludes by asserting that future AI program must:
"model [...] our particular way of being embedded and embodied such that what we experience is significant for us in the particular way that it is [...] with our needs, desires, pleasures, pains, ways of moving, cultural background, etc."
In other words, AI must also model the motivations that guide human activity. However, insofar as Dreyfus's critique assumes that artificial intelligence must be evaluated against the standards of human intelligence, recent proposals to reorient toward "augmented intelligence" may suggest an alternative path forward.
.