Chat GPT seems like a those RAs that sometimes show up in 1st year grad programs--indifferent or willingly lying to get the job completed while minimizing effort. Test: have Chat search for some obscure data or technical, non-literary fact. It's not unusual for Chat to say the data doesn't exist or the fact isn't there. You can insist the data or fact is there, and sometime Chat will find it and apologize. So, be careful accepting Chat's factual claims--it's very lazy as a researcher. In contrast, it's helpful as an editor of prewritten text and amusingly successful in writing poetic forms and laudatory text for birthday and retirement congratulations. Being lazy, low cost and convincing, Chat and AI will dominate mass media, corporate and government writing.
Agreed: the more practical implications at this time (significant job displacement or adjustment, the ability to create fake videos etc) are the things I worry about most though these do not at all call for a pause.
I’m far from an expert on this topic but to the extent these models are trained on data that is ultimately from humans, it does seem that there should be an asymptote.
To me the asymptote of anything that is trained on human data is becoming the smartest person in every topic and having incredible speed at that. But I don’t see how that enables some AI take over.
Thus far, Chatgpt is an animated Macauley's schoolboy. I'm not sure we will even recognize intelligence if we see it; our understanding is too crude: if a monkey bites you, that's instinct, if it smacks you with a 2x4, that's intelligence.
I'd guess LLMs might be most scary as a disruptor. How will this change our lives? That's not an unimportant question but its way different than AI taking over. AI will probably get scarier when it starts controlling things. ... Oh yeah, driverless cars. Well, surely the next advance will lead to catastrophe.
I'm still trying to figure out the purpose or value of a six month pause. And why is GPT-4 anything but an arbitrary upper limit on "power." And why would anyone think they could get a pause in development? Maybe the request could delay putting something new on the market but even that seems a stretch. Do they want government to legislate AI development restrictions?
I am not a fan of a pause because these models pose almost 0 extinction risk. I am using ChatGPT4 to review my Spanish essays and to correct and grammar or usage errors. As compared to the corrections of my teacher ChatGPT4 gets a 3 out of 10. It is a gigantic jump to go from here to HAL 9000 and on that long journey we can access catastrophic risks.
In addition the really hard problems of in innovations take lots of data and there does not seem to be enough data nor is there ever likely to be enough data for this.
Nevertheless, LLM’s are invaluable resources and should be used to their fullest without impediment.
It seems the issues with hallucinations and other downsides to ChatGPT-4 suggest its rather useful for society for them to continue training better models to fix some of the issues. Its unclear what the point of a pause would be, other than for Elon Musk who has suggested he may want to do a non-woke AI, to have time to get his act together to compete. It seems a case of "Bootleggers and Baptists": those truly scared about AI and those that just want to hold back their competitors.
For those scared of AGI: since these systems aren't anywhere close to that, its unclear what the point of pausing them is. Many like AI pioneer and head of AI at Meta Yann LeCun see the current LLMs as very useful, but a detour from any path to actual AGI. It would seem those scared of true AGI should want resources going to the detour and distracting researchers from spending time on a path towards AGI.
It seems likely only those already concerned about AI safety would pause, while those who weren't would continue on. That would undermine AIs concerned about safety's ability to compete in the marketplace and be detrimental to their goals. Except its unlikely there are any such places, except those not close to being able to train a GPT-5 level model, who might I suppose pretend they were holding off on doing so to get a PR boost when they finally are ready to do so at some point 6+ months down the road as if they were being "safer" in hopes that gets them customers.
It seems like for most they likely didn't expect the pause to happen but figured it was a good publicity stunt to try to scare politicians and news media into covering AI risks. They saw all the coverage covid doomsayers got, and climate change alarmists.
re: Bing AI being terse in its response: it depends on what mode its in, they have 3: creative, balanced, and precise. Its underlying AI is OpenAI's GPT-4 model, though its unclear what training differences there may be on top of it. Creative is more likely to give more verbose responses: but regardless you can always tell it to give a more detailed or longer response. Its response to me in creative mode for "write about David Hackett-Fischer’s Albion’s Seed" wasn't as brief as you quoted (in it the bullet points were formatted with bullet points that don't show up here, and it linked to sources but many comment systems moderate lots of links so left those out):
"Here is a brief summary of David Hackett-Fischer’s Albion’s Seed:
Albion’s Seed is a book that explores the cultural origins and diversity of the United States by tracing the folkways of four groups of British immigrants who settled in different regions of North America from 1629 to 1775. The author argues that these folkways, which include language, religion, family, customs, politics and values, have persisted and influenced the American society and identity for centuries. The four groups are:
The Puritans, who came from East Anglia to Massachusetts and established a religious and communal society based on moral order and education.
The Cavaliers, who came from the South of England to Virginia and established a hierarchical and aristocratic society based on honor and slavery.
The Quakers, who came from the North Midlands and Wales to the Delaware Valley and established a tolerant and egalitarian society based on peace and cooperation.
The Borderers, who came from the Borderlands of North Britain and Ireland to the Backcountry and established a violent and individualistic society based on liberty and resistance.
The book also examines how these regional cultures have shaped the American politics, economy, law and culture over time, and how they have contributed to the pluralism and democracy of the United States. Albion’s Seed is a comprehensive and fascinating account of the British roots of American diversity1"
Chat GPT seems like a those RAs that sometimes show up in 1st year grad programs--indifferent or willingly lying to get the job completed while minimizing effort. Test: have Chat search for some obscure data or technical, non-literary fact. It's not unusual for Chat to say the data doesn't exist or the fact isn't there. You can insist the data or fact is there, and sometime Chat will find it and apologize. So, be careful accepting Chat's factual claims--it's very lazy as a researcher. In contrast, it's helpful as an editor of prewritten text and amusingly successful in writing poetic forms and laudatory text for birthday and retirement congratulations. Being lazy, low cost and convincing, Chat and AI will dominate mass media, corporate and government writing.
Agreed: the more practical implications at this time (significant job displacement or adjustment, the ability to create fake videos etc) are the things I worry about most though these do not at all call for a pause.
I’m far from an expert on this topic but to the extent these models are trained on data that is ultimately from humans, it does seem that there should be an asymptote.
To me the asymptote of anything that is trained on human data is becoming the smartest person in every topic and having incredible speed at that. But I don’t see how that enables some AI take over.
It is only a matter of time before someone publishes a wildly successful novel written by ChatGPT. Will Open AI sue for copyrights?
Thus far, Chatgpt is an animated Macauley's schoolboy. I'm not sure we will even recognize intelligence if we see it; our understanding is too crude: if a monkey bites you, that's instinct, if it smacks you with a 2x4, that's intelligence.
Agreed.
I'd guess LLMs might be most scary as a disruptor. How will this change our lives? That's not an unimportant question but its way different than AI taking over. AI will probably get scarier when it starts controlling things. ... Oh yeah, driverless cars. Well, surely the next advance will lead to catastrophe.
I'm still trying to figure out the purpose or value of a six month pause. And why is GPT-4 anything but an arbitrary upper limit on "power." And why would anyone think they could get a pause in development? Maybe the request could delay putting something new on the market but even that seems a stretch. Do they want government to legislate AI development restrictions?
I am not a fan of a pause because these models pose almost 0 extinction risk. I am using ChatGPT4 to review my Spanish essays and to correct and grammar or usage errors. As compared to the corrections of my teacher ChatGPT4 gets a 3 out of 10. It is a gigantic jump to go from here to HAL 9000 and on that long journey we can access catastrophic risks.
In addition the really hard problems of in innovations take lots of data and there does not seem to be enough data nor is there ever likely to be enough data for this.
Nevertheless, LLM’s are invaluable resources and should be used to their fullest without impediment.
It seems the issues with hallucinations and other downsides to ChatGPT-4 suggest its rather useful for society for them to continue training better models to fix some of the issues. Its unclear what the point of a pause would be, other than for Elon Musk who has suggested he may want to do a non-woke AI, to have time to get his act together to compete. It seems a case of "Bootleggers and Baptists": those truly scared about AI and those that just want to hold back their competitors.
For those scared of AGI: since these systems aren't anywhere close to that, its unclear what the point of pausing them is. Many like AI pioneer and head of AI at Meta Yann LeCun see the current LLMs as very useful, but a detour from any path to actual AGI. It would seem those scared of true AGI should want resources going to the detour and distracting researchers from spending time on a path towards AGI.
It seems likely only those already concerned about AI safety would pause, while those who weren't would continue on. That would undermine AIs concerned about safety's ability to compete in the marketplace and be detrimental to their goals. Except its unlikely there are any such places, except those not close to being able to train a GPT-5 level model, who might I suppose pretend they were holding off on doing so to get a PR boost when they finally are ready to do so at some point 6+ months down the road as if they were being "safer" in hopes that gets them customers.
It seems like for most they likely didn't expect the pause to happen but figured it was a good publicity stunt to try to scare politicians and news media into covering AI risks. They saw all the coverage covid doomsayers got, and climate change alarmists.
re: Bing AI being terse in its response: it depends on what mode its in, they have 3: creative, balanced, and precise. Its underlying AI is OpenAI's GPT-4 model, though its unclear what training differences there may be on top of it. Creative is more likely to give more verbose responses: but regardless you can always tell it to give a more detailed or longer response. Its response to me in creative mode for "write about David Hackett-Fischer’s Albion’s Seed" wasn't as brief as you quoted (in it the bullet points were formatted with bullet points that don't show up here, and it linked to sources but many comment systems moderate lots of links so left those out):
"Here is a brief summary of David Hackett-Fischer’s Albion’s Seed:
Albion’s Seed is a book that explores the cultural origins and diversity of the United States by tracing the folkways of four groups of British immigrants who settled in different regions of North America from 1629 to 1775. The author argues that these folkways, which include language, religion, family, customs, politics and values, have persisted and influenced the American society and identity for centuries. The four groups are:
The Puritans, who came from East Anglia to Massachusetts and established a religious and communal society based on moral order and education.
The Cavaliers, who came from the South of England to Virginia and established a hierarchical and aristocratic society based on honor and slavery.
The Quakers, who came from the North Midlands and Wales to the Delaware Valley and established a tolerant and egalitarian society based on peace and cooperation.
The Borderers, who came from the Borderlands of North Britain and Ireland to the Backcountry and established a violent and individualistic society based on liberty and resistance.
The book also examines how these regional cultures have shaped the American politics, economy, law and culture over time, and how they have contributed to the pluralism and democracy of the United States. Albion’s Seed is a comprehensive and fascinating account of the British roots of American diversity1"