What if we design a system in which a LLM generates the code and training data for a new generation of LLM which generates the code and training data for the next? Is it possible that we see them spiraling fast to the best LLM possible?
GPT-4
271–280 of 1001 posts
Re: GPT-4
#272A class of problem that GPT-4 appears to still really struggle with is variants of common puzzles. For example: >Suppose I have a cabbage, a goat and a lion, and I need to get them across a river. I have a boat that can only carry myself and a single other item. I am not allowed to leave the cabbage and lion alone together, and I am not allowed to leave the lion and goat alone together. How can I safely get all three…
I think you may have misstated the puzzle. It's ok to leave the lion and the cabbage together, assuming it's not a vegetarian lion.
Re: GPT-4
#273A class of problem that GPT-4 appears to still really struggle with is variants of common puzzles. For example: >Suppose I have a cabbage, a goat and a lion, and I need to get them across a river. I have a boat that can only carry myself and a single other item. I am not allowed to leave the cabbage and lion alone together, and I am not allowed to leave the lion and goat alone together. How can I safely get all three…
the "trick" Monty Hall problems are another good one here: https://twitter.com/colin_fraser/status/1628461980645462016
Apparently GPT-4 gets this one right!
Re: GPT-4
#274Test taking will change. In the future I could see the student engaging in a conversation with an AI and the AI producing an evaluation. This conversation may be focused on a single subject, or more likely range over many fields and ideas. And may stretch out over months. Eventually teaching and scoring could also be integrated as the AI becomes a life-long tutor. Even in a future where human testing/learning is no l…
I think a mass market version of the young lady’s illustrated primer from Neal Stephenson’s Diamond Age would so deeply transform society as to make it unrecognizable, and the way things are going that product is a few years away.
I’m really questioning what to do about this professionally, because it is obvious this technology will radically reshape my job, but it is unclear how.
Re: GPT-4
#275Today's submarine from the NYT is hilariously unintentionally unflattering, did someone even proof read this? https://www.nytimes.com/2023/03/14/technology/openai-new-gpt... Mnemonics to learn basic words: Hola:"Hola, cola" - sounds like "cola" (Coke) Adiós: "Adiós, my toes" - rhymes with "toes" Gracias: "Gracias, grassy ass" - sounds like "grassy ass" Mañana: "Mañana, banana" - rhymes with "banana" Uno: "Uno, you kn…
> Mañana: "Mañana, banana" - rhymes with "banana"
Uh... Almost?
> Gracias: "Gracias, grassy ass" - sounds like "grassy ass"
This seems like it would encourage very coarse pronunciation with an obnoxious accent.
Overall this appears relatively bad. I'm not sure why they would want this to be publicly visible yet.
Re: GPT-4
#276From the paper: > Given both the competitive landscape and the safety implications of large-scale models like GPT-4, this report contains no further details about the architecture (including model size), hardware, training compute, dataset construction, training method, or similar. I'm curious whether they have continued to scale up model size/compute significantly or if they have managed to make significant innovati…
I bet they use CLIP to caption the image and feed the text of the caption into GPT, but that's just a guess.
There's no need to round-trip through text, you "just" need to train an embedding space that captures both domains.
Re: GPT-4
#277A class of problem that GPT-4 appears to still really struggle with is variants of common puzzles. For example: >Suppose I have a cabbage, a goat and a lion, and I need to get them across a river. I have a boat that can only carry myself and a single other item. I am not allowed to leave the cabbage and lion alone together, and I am not allowed to leave the lion and goat alone together. How can I safely get all three…
For anyone wondering you bring the lion across. The trick is that it’s the lion that eats the cabbage not the goat.
Re: GPT-4
#278Most of the comments here are denial and goalpost shifting... GPT-4 has different strengths and weaknesses from humans, but it is now in the general realm of human intelligence vs being far below that with GPT-3. Another jump past GPT-4 of the same magnitude, would greatly surpass human cognitive abilities and present a danger to humanity.
Thank you.
Every single step forward with AI is met with a massive amount of people shrugging it off for whatever latest goal post they plant.
Re: GPT-4
#279I think it's interesting that they've benchmarked it against an array of standardized tests. Seems like LLMs would be particularly well suited to this kind of test by virtue of it being simple prompt:response, but I have to say...those results are terrifying. Especially when considering the rate of improvement. bottom 10% to top 10% of LSAT in What are the implications for society when general thinking, reading, and…
> bottom 10% to top 10% of LSAT in People still really find it hard to internalize exponential improvement.
So many evaluations of LLMs were saying things like "Don't worry, your job is safe, it still can't do X and Y."
My immediate thought was always, "Yes, the current version can't, but what about a few weeks or months from now?"
Re: GPT-4
#280A class of problem that GPT-4 appears to still really struggle with is variants of common puzzles. For example: >Suppose I have a cabbage, a goat and a lion, and I need to get them across a river. I have a boat that can only carry myself and a single other item. I am not allowed to leave the cabbage and lion alone together, and I am not allowed to leave the lion and goat alone together. How can I safely get all three…
I think you may have misstated the puzzle. It's ok to leave the lion and the cabbage together, assuming it's not a vegetarian lion.
It’s an intensionally misstated problem to see if gpt4 will correctly take the lion across first and leave the goat.