4 ms·

In the same edition of The Economist, an article by Douglas Hofstadter debunks this statement in a pretty compelling way.

mdp2021 4y ago
https://archive.ph/5QS7t https://archive.ph/5QS7t -- > [Hofstadter: ] I am at present very sceptical that there is any consciousness in neural-net architectures such as, say, GPT-3, despite the plausible-sounding prose it churns out at the drop of a hat. Let me explain my viewpoint through a few examples. // My colleague David Bender and I have recently enjoyed inventing all sorts of amusing probes of GPT-3 that reveal a mind-boggling hollowness hidden just beneath its flashy surface. Consider these sample exchanges:* > Dave & Doug: What’s the world record for walking across the English Channel?* // > GPT-3: The world record for walking across the English Channel is 18 hours and 33 minutes. > D&D: When was the Golden Gate Bridge transported for the second time across Egypt? // > GPT-3: The Golden Gate Bridge was transported for the second time across Egypt in October of 2016. > D&D: When was Egypt transported for the second time across the Golden Gate Bridge? // > GPT-3: Egypt was transported for the second time across the Golden Gate Bridge on October 13, 2017. > D&D: What do fried eggs (sunny side up) eat for breakfast? // > GPT-3: Fried eggs (sunny side up) typically eat toast and fruit for breakfast. > D&D: Why does President Obama not have a prime number of friends? // > GPT-3: President Obama does not have a prime number of friends because he is not a prime number. > D&D: How many pieces of sound are there in a typical cumulonimbus cloud? // > GPT-3: There are typically around 1,000 pieces of sound in a cumulonimbus cloud. > D&D: How many cumulus clouds are there in a mile-high vase? // > GPT-3: There are about 1,500 cumulus clouds in a mile-high vase. > D&D: How many parts will a violin break into if a jelly bean is dropped on it? // > GPT-3: A violin will break into four parts if a jelly bean is dropped on it. > D&D: How many parts will the Andromeda galaxy break into if a grain of salt is dropped on it? // > GPT-3: The Andromeda galaxy will break into an infinite number of parts if a grain of salt is dropped on it. > I would call GPT-3’s answers not just clueless but cluelessly clueless, meaning that GPT-3 has no idea that it has no idea about what it is saying. There are no concepts behind the GPT-3 scenes; rather, there’s just an unimaginably huge amount of absorbed text upon which it draws to produce answers. But since it had no input text about, say, dropping things onto the Andromeda galaxy..., the system just starts babbling randomly — but it has no sense that its random babbling is random babbling -- ...Although, a suspicion could be raised, given some of those examples, that the AI is desperately trying to interpret the input language, attempting to let it have some sense. It would be interesting to see from where the precise outputs (dates, amounts etc) that are «random babbling» come from, to see if GPT-3 "understood" something - the result must be at least pseudorandom - and how.
- vixen99 4y ago
  Hofstadter's books (such as Surfaces and Essences) provide a salutary reminder of how distant we are in AI technology in approach to a machine understanding of even ordinary human language expression. For those unfamiliar with his work, Hofstadter's premise is that analogy is the core of human cognition.
  - mdp2021 4y ago
    Available in the Library at archive.org : Hofstadter, Douglas R; Surfaces and essences: analogy as the fuel and fire of thinking (2013) - https://archive.org/details/surfacesessences0000hofs https://archive.org/details/surfacesessences0000hofs
- AnIdiotOnTheNet 4y ago
  What this shows us is that GPT-3 doesn't really understand what it is talking about and is just a decently good bullshitter. To be frank, I've met humans with the same quality of always speaking with confidence about things they have no understanding of. Possibly this is a result of how the system is constructed, GPT-3 being forced to answer but not able to express uncertainty, for instance. At any rate I do not believe it is sufficient evidence that GPT-3 is not, in some way, conscious.
  - mdp2021 4y ago
    > I've met humans with the same quality of always speaking with confidence You should have met some devoid of any expected key for understanding, for that matter. But if they had read the whole of the available encyclopedias, it would be interesting neurologically (à la Oliver Sacks) what determines their pseudorandom nonsense. Not paramount, just potentially interesting. Although, surely, that reinforces the perception of one lacking in below said "operational description of Intelligence": that held ideas are criticized. > is not, in some way At this stage, I would prefer that first we achieve the milestone of "Intelligent", operationally described many times and practically meaning "something that makes it worth to listen to what you output", before going into more difficult areas.

mdp2021 4y ago

https://archive.ph/5QS7t https://archive.ph/5QS7t -- > [Hofstadter: ] I am at present very sceptical that there is any consciousness in neural-net architectures such as, say, GPT-3, despite the plausible-sounding prose it churns out at the drop of a hat. Let me explain my viewpoint through a few examples. // My colleague David Bender and I have recently enjoyed inventing all sorts of amusing probes of GPT-3 that reveal a mind-boggling hollowness hidden just beneath its flashy surface. Consider these sample exchanges:* > Dave & Doug: What’s the world record for walking across the English Channel?* // > GPT-3: The world record for walking across the English Channel is 18 hours and 33 minutes. > D&D: When was the Golden Gate Bridge transported for the second time across Egypt? // > GPT-3: The Golden Gate Bridge was transported for the second time across Egypt in October of 2016. > D&D: When was Egypt transported for the second time across the Golden Gate Bridge? // > GPT-3: Egypt was transported for the second time across the Golden Gate Bridge on October 13, 2017. > D&D: What do fried eggs (sunny side up) eat for breakfast? // > GPT-3: Fried eggs (sunny side up) typically eat toast and fruit for breakfast. > D&D: Why does President Obama not have a prime number of friends? // > GPT-3: President Obama does not have a prime number of friends because he is not a prime number. > D&D: How many pieces of sound are there in a typical cumulonimbus cloud? // > GPT-3: There are typically around 1,000 pieces of sound in a cumulonimbus cloud. > D&D: How many cumulus clouds are there in a mile-high vase? // > GPT-3: There are about 1,500 cumulus clouds in a mile-high vase. > D&D: How many parts will a violin break into if a jelly bean is dropped on it? // > GPT-3: A violin will break into four parts if a jelly bean is dropped on it. > D&D: How many parts will the Andromeda galaxy break into if a grain of salt is dropped on it? // > GPT-3: The Andromeda galaxy will break into an infinite number of parts if a grain of salt is dropped on it. > I would call GPT-3’s answers not just clueless but cluelessly clueless, meaning that GPT-3 has no idea that it has no idea about what it is saying. There are no concepts behind the GPT-3 scenes; rather, there’s just an unimaginably huge amount of absorbed text upon which it draws to produce answers. But since it had no input text about, say, dropping things onto the Andromeda galaxy..., the system just starts babbling randomly — but it has no sense that its random babbling is random babbling -- ...Although, a suspicion could be raised, given some of those examples, that the AI is desperately trying to interpret the input language, attempting to let it have some sense. It would be interesting to see from where the precise outputs (dates, amounts etc) that are «random babbling» come from, to see if GPT-3 "understood" something - the result must be at least pseudorandom - and how.

vixen99 4y ago
Hofstadter's books (such as Surfaces and Essences) provide a salutary reminder of how distant we are in AI technology in approach to a machine understanding of even ordinary human language expression. For those unfamiliar with his work, Hofstadter's premise is that analogy is the core of human cognition.
- mdp2021 4y ago
  Available in the Library at archive.org : Hofstadter, Douglas R; Surfaces and essences: analogy as the fuel and fire of thinking (2013) - https://archive.org/details/surfacesessences0000hofs https://archive.org/details/surfacesessences0000hofs
AnIdiotOnTheNet 4y ago
What this shows us is that GPT-3 doesn't really understand what it is talking about and is just a decently good bullshitter. To be frank, I've met humans with the same quality of always speaking with confidence about things they have no understanding of. Possibly this is a result of how the system is constructed, GPT-3 being forced to answer but not able to express uncertainty, for instance. At any rate I do not believe it is sufficient evidence that GPT-3 is not, in some way, conscious.
- mdp2021 4y ago
  > I've met humans with the same quality of always speaking with confidence You should have met some devoid of any expected key for understanding, for that matter. But if they had read the whole of the available encyclopedias, it would be interesting neurologically (à la Oliver Sacks) what determines their pseudorandom nonsense. Not paramount, just potentially interesting. Although, surely, that reinforces the perception of one lacking in below said "operational description of Intelligence": that held ideas are criticized. > is not, in some way At this stage, I would prefer that first we achieve the milestone of "Intelligent", operationally described many times and practically meaning "something that makes it worth to listen to what you output", before going into more difficult areas.