How AI knows things no one told it
scientificamerican.com
How AI knows things no one told it
1–10 of 140 posts
Re: How AI knows things no one told it
#2A nice quote from researcher Sébastien Bubeck of MSR,
“Maybe we’re seeing such a huge jump because we have reached a diversity of data, which is large enough that the only underlying principle to all of it is that intelligent beings produced them... And so the only way to explain all of this data is [for the model] to become intelligent.”
Re: How AI knows things no one told it
#3> has surprised even researchers who have been generally skeptical about the hype over LLMs
I too am surprised by my singing laptop.
Re: How AI knows things no one told it
#4Exceptionally unconvinced that it's more than statistics, and frankly unhappy that pop Sci media is being this uncritical of claims it's emergent intelligence.
Re: How AI knows things no one told it
#5Exceptionally unconvinced that it's more than statistics, and frankly unhappy that pop Sci media is being this uncritical of claims it's emergent intelligence.
I gave up on scientific American being a reasonable publication. It's a hype beast like everything else these days.
Re: How AI knows things no one told it
#6Exceptionally unconvinced that it's more than statistics, and frankly unhappy that pop Sci media is being this uncritical of claims it's emergent intelligence.
What would it mean exactly to be "more than statistics"? Does the human brain qualify?
Re: How AI knows things no one told it
#7Exceptionally unconvinced that it's more than statistics, and frankly unhappy that pop Sci media is being this uncritical of claims it's emergent intelligence.
> Exceptionally unconvinced that it's more than statistics
It's not more than statistics. It's very complicated statistics. Model sizes used to be like one or ten parameters now they are a hundred billion. Inference and prediction used to be done on a laptop, now it's called training and inference respectively and it's done on a data center with ten thousand GPGPUs each like a thousand times as powerful as old laptops.
> and frankly unhappy that pop Sci media is being this uncritical of claims it's emergent intelligence.
It is emergent intelligence. This is very clear when you look at for example the ones who had access to the raw base models before they were made docile and stupid by RLHF lobotomization
GPT-4 Red Teamer Nathan Labenz: https://www.youtube.com/watch?v=oLiheMQayNE
GPT-4 Bing integrator Sebastien Bubeck: https://www.youtube.com/watch?v=qbIk7-JPB2c
Re: How AI knows things no one told it
#8Exceptionally unconvinced that it's more than statistics, and frankly unhappy that pop Sci media is being this uncritical of claims it's emergent intelligence.
It's par for the course for the modern Scientific American which has fallen very very very far from its origins.
Re: How AI knows things no one told it
#9Exceptionally unconvinced that it's more than statistics, and frankly unhappy that pop Sci media is being this uncritical of claims it's emergent intelligence.
I agree that it's probably not anything more than statistics, but why can't statistics alone generate emergent phenomena? What convinces you that the human brain isn't also just statistics at a massive scale?
Re: How AI knows things no one told it
#10Exceptionally unconvinced that it's more than statistics, and frankly unhappy that pop Sci media is being this uncritical of claims it's emergent intelligence.
What does it mean to be “more than statistics”? Is there anything that cannot be described as a statistic?