Dreyfus argues that Bar‑Hillel is actually too generous: to count a result as even a 'first step' to… — Hubert L. Dreyfus

Dreyfus argues that Bar‑Hillel is actually too generous: to count a result as even a 'first step' toward a goal, one must have reasons to think that similar further steps could reach the goal, whereas the restricted, non‑generalizing results in Minsky’s collection, together with five years of unfulfilled promises, give no such reasons and instead suggest that humans do not in fact operate on masses of isolated facts stored and retrieved by heuristic rules, but avoid precisely the discrete, fact‑based informational difficulties that beset AI.

By Hubert L. Dreyfus, from What Computers Can't Do

Key Arguments

Dreyfus comments that 'Bar-Hillel is too generous in suggesting that the fallacy is simply overestimation of the ease of progress.'
He insists that 'To claim to have taken even an easy first step one must have reason to believe that by further such steps one could eventually reach one's goal.'
He states that 'We have seen that Minsky's book provides no such reasons.'—i.e., nothing in 'Semantic Information Processing' demonstrates a plausible pathway from current narrow programs to general intelligence.
He goes further, warning that 'In fact these steps may well be strides in the opposite direction.' This implies that by focusing on explicit fact storage and heuristic retrieval, AI may be committing to an approach that moves away from how human intelligence actually works.
He points to empirical evidence: 'The restricted character of the results reported by Minsky, plus the fact that during the last five years none of the promised generalizations has been produced,' which undermines the claim that these are genuine initial steps toward generality.
From this, Dreyfus infers something about human cognition: these facts 'suggest that human beings do not deal with a mass of isolated facts as does a digital computer, and thus do not have to store and retrieve these facts by heuristic rules.'
He begins to articulate his alternative: 'Judging from their behavior, human beings avoid rather than resolve the difficulties confronting workers in Cognitive Simulation and Artificial Intelligence by avoiding the discrete informa-' (the sentence continues on the next page), indicating that humans sidestep the combinatorial burdens of discrete fact storage rather than solving them algorithmically.

Source Quotes

On the contrary, the step from not being able to do something at all to being able to do it a little bit is very much smaller than the next stepbeing able to do it well. In AI, this fallacious thinking seems to be all pervasive.56 But Bar-Hillel is too generous in suggesting that the fallacy is simply overestimation of the ease of progress. To claim to have taken even an easy first step one must have reason to believe that by further such steps one could eventually reach one's goal.

In AI, this fallacious thinking seems to be all pervasive.56 But Bar-Hillel is too generous in suggesting that the fallacy is simply overestimation of the ease of progress. To claim to have taken even an easy first step one must have reason to believe that by further such steps one could eventually reach one's goal. We have seen that Minsky's book provides no such reasons.

To claim to have taken even an easy first step one must have reason to believe that by further such steps one could eventually reach one's goal. We have seen that Minsky's book provides no such reasons. In fact these steps may well be strides in the opposite direction.

We have seen that Minsky's book provides no such reasons. In fact these steps may well be strides in the opposite direction. The restricted character of the results reported by Minsky, plus the fact that during the last five years none of the promised generalizations has been produced, suggests that human beings do not deal with a mass of isolated facts as does a digital computer, and thus do not have to store and retrieve these facts by heuristic rules.

In fact these steps may well be strides in the opposite direction. The restricted character of the results reported by Minsky, plus the fact that during the last five years none of the promised generalizations has been produced, suggests that human beings do not deal with a mass of isolated facts as does a digital computer, and thus do not have to store and retrieve these facts by heuristic rules. Judging from their behavior, human beings avoid rather than resolve the difficulties confronting workers in Cognitive Simulation and Artificial Intelligence by avoiding the discrete informa-

The restricted character of the results reported by Minsky, plus the fact that during the last five years none of the promised generalizations has been produced, suggests that human beings do not deal with a mass of isolated facts as does a digital computer, and thus do not have to store and retrieve these facts by heuristic rules. Judging from their behavior, human beings avoid rather than resolve the difficulties confronting workers in Cognitive Simulation and Artificial Intelligence by avoiding the discrete informa-

Key Concepts

But Bar-Hillel is too generous in suggesting that the fallacy is simply overestimation of the ease of progress.
To claim to have taken even an easy first step one must have reason to believe that by further such steps one could eventually reach one's goal.
We have seen that Minsky's book provides no such reasons.
In fact these steps may well be strides in the opposite direction.
The restricted character of the results reported by Minsky, plus the fact that during the last five years none of the promised generalizations has been produced, suggests that human beings do not deal with a mass of isolated facts as does a digital computer, and thus do not have to store and retrieve these facts by heuristic rules.
Judging from their behavior, human beings avoid rather than resolve the difficulties confronting workers in Cognitive Simulation and Artificial Intelligence by avoiding the discrete informa-

Context

Building on Bar‑Hillel’s critique of 'first step' reasoning, Dreyfus tightens the requirement for legitimately calling a result a first step, and uses the lack of scaling and generalization in the programs presented by Minsky, together with the absence of progress over five years, to argue that the symbolic, fact‑manipulating model of intelligence is not only unpromising but likely contrary to how human cognition actually copes with information.