GPT4o still can't do the intersection of two different ideas that are not in the training set. It can't even produce random variations on the intersection of two different ideas.
Further though, we shouldn't expect the model to do this. It is not fair to the model and its actual usefulness and how amazing what the models can do with zero understanding. To believe the model understands is to fool yourself.
I am blown away having spent hours prompting GPT4o.
If it can give shorter answers in voice mode instead of lectures then a back and forth conversation with this much power can be quite interesting.
I still doubt I would use it that much though just because of how much is lost compared to the screen. Code and voice make no sense. The time between prompts usually requires quite a bit of thought for anything interesting that a conversation itself is only useful for things I have already asked it.
For me, gpt4 is already as useless as 3.5. I will never prompt gpt4 again. I can still push GPT4o over the edge in python but damn, it is pretty out there. Then the speed is really amazing.
I wasn't impressed in the first 5 minutes of using it but it is quite impressive after 2 solid hours of random topics.
Much faster for sure but I have also not had anything give an error in python with jupyter. Usually you could only stray so far with more obscure python libraries before it starts producing errors.
That much better than 4 in chess is pretty shocking in a great way.
While I don't agree at all with you, I very much appreciate reading something like this that I don't agree at all with. This to me encapsulates the beauty of human interaction.
It is exactly what will be missing from language model interaction. I don't want something that agrees with me and I don't want something that is pretending to randomly disagree with me either.
The fun of this interaction is maybe one of us flips the other to their point of view.
I can completely picture how to take the HN API and the chatGPT API to make my own personal HN to post on and be king of the castle. Everyone can just upvote my responses to prove what a genius I am. That obviously would be no fun. There is no fun configuration of that app though either with random disagreements and algorithmic different points of view.
I think you can pretty much apply that to all domains of human interaction that is not based on pure information transfer.
There is a reason we are a year in and the best we can do are new stories about someone making X amount of money with their AI girlfriend and follow up new about how its the doom of society. It has nothing to do with reality.
I don't know, my experience is that it is very hard to tell if the model is better or worse with an update.
One day I will have an amazing session and the next it seems like it has been nerfed only to give better results than ever the next day. Wash, rinse , repeat and randomize that ordering.
So far, I would have not be able to tell the difference between 4 and 4o.
If this is the new 3.5 though then 5 will be worth the wait to say the least.
People want to believe in something more when it comes to ancient Egypt even if they know it is not reality. To me, it is obviously linked to the beauty and other worldly feel of the ancient Egyptian iconography and art.
If you just had the pyramids in the desert with the entire civilization lost, while impressive there wouldn't be other worldly explanations that sounded palatable. It would obviously just be the engineering of a lost civilization.
It is really a testament to the power of ancient Egyptian art that it can still inspire the imagination to such a degree thousands of years later and across cultures.
I think for some that it is the opposite. Most people are able to understand that due to the extensive evidence of centuries of advanced civilization, including extensive surviving writings and evidence of mathematics knowledge, the Egyptians were able to coordinate the construction of the pyramids using advanced, but not anachronistic technology. However, due to the disproportionate vastness and majesty of the pyramids and some of the other surviving ancient Egyptian works, many people can still be led to believe that there was something more at work than practical engineering with ancient technology.
The real problem though are the ridiculous cost of state run schools.
As someone who has been out of college for 20 years, I can't believe what my state school charges now.
When I graduated my school was a good value compared to private options. That was the whole point of a state school.
Now the same school cost 4X what it does and I don't know if I could even get in when up against so many brilliant Indian and Chinese students. I assumed when I went the goal of the institution was for a more educated population in the state. That was the point of the huge difference in price for instate tuition.
Now it seems like some kind of money making racket.
Afaik, state schools have lost a lot of budget from states (in inflation-adjusted terms) compared to pre-90s.
Consequently, they balance the books by taking on more foreign/out-of-state students (who they can charge much more).
At the core, though, it appears to be a supply and demand problem: there's a ton of demand for people who want to go to college, colleges don't have the resources (or interest?) to expand supply, and so they raise prices because they can.
State schools have also been wasting a fortune on gold-plated facilities. I was just at UCSD yesterday and it was ridiculously nice. This is not a responsible use of tuition and tax dollars. We need an austerity program. If that makes the state schools less attractive to top students and professors then so be it.
This is moreso a problem with government/not-for-profit budgetary processes and capex vs opex.
With government funding, it's use-it-or-lose-it. I.e. state legislature will definitely allocate it to something else next year, if they come in under-budget in a year.
Add on top of that that state schools have the NASA problem: they're effectively performing (to state legislature and donors) for their dinner. And big, splashy buildings (with dedication names!) impress both. And attract picky faculty and students.
With not-for-profit, facilities are one of the few allowed places you can dump excess cash.
I floated in another comment that government grant/loan programs should impose an ACA-style admin vs teaching cost cap on institutions for participation. Maybe for capex too?
The world knows this is art but some bullshit artists, pun intended, in New York pretend like it is not. Then other bullshit artists in other cities follow what the bullshit artists in New York are doing because most aren't creative or free thinkers at all.
I love galleries personally but it is a class of non-creative, closed minded, bullshit artists at this point.
Also because those bullshit artists are aware the big money is in this bullshit world, so they give their best to get in and profit. You certainly don't get rich drawing for Bethesda, while in the art gallery business you maybe maybe maybe could.
GPT4o still can't do the intersection of two different ideas that are not in the training set. It can't even produce random variations on the intersection of two different ideas.
Further though, we shouldn't expect the model to do this. It is not fair to the model and its actual usefulness and how amazing what the models can do with zero understanding. To believe the model understands is to fool yourself.