(Grants do not require repayment.)
You paint the picture as if the All or nothing folks are extreme when in reality the extreme is more likely then some perfectly fair and balanced equilibrium.
In nature things tend to overload or fizzle out or stay in equilibrium. Equilibrium, though possible, is the rarer outcome. Mind you it's not an impossible outcome but given the way entropy works, it is the rarer outcome.
Humanity itself is an example of this rare outcome. Usually molecules don't self assemble into replicating machines, they either freeze into inanimate rock or overload into fusion producing stars.
As for AI. I think it will either change the world, or amount to nothing. The former seems more likely. Some strange middle ground where the ai technology never improves to some point of a societal paradigm shift seems unlikely. chatGPT and sora only makes me ask what is the trend line predicting next?
What I find is interesting, is that the negative news regarding AI safety is adding to the hype as well since it seems to capture a lot of attention.
The people that try the models and find no value will obviously not continue to use something they get no value from.
If you discount anyone that doesn’t use it on a continual basis, then of course all the detractors aren’t as heavy of users as those that do use it regularly.
I think a better argument would be the size of the set of regular users vs. some grounded guesstimate for how many have tried it.
I would say for AI deniers it's the losers fallacy. You guys haven't taken the time to thoroughly examine chatGPT4. You just stop at the first hallucination. It's a paradigm shifting AI.
Detractors like you just rely on common tropes like oh "stop anthropomorphizing it" and stuff like that. People who see this stuff as game changing recognize these biases and we still say it's game changing. Nobody is stupid enough to think this thing is even remotely human.
I asked chatGPT to program me a complex GUI in Python about selecting multiple time segments out of an overall interval using sliders. It gave me working python code.
You have realize that this thing has no eyes and it was able to program a visual interface as if it did. I'm not anthropomorphizing anything when I say this, but on some level chatGPT "understands" what you are giving it as a query.
Ah, been a while since someone pigeon-holed me like that. Usually it's the opposite pigeon hole I get stuffed into.
Ironically, that kind of rapid but inaccurate misreading is one of the things that makes LLMs somewhat easier to deal with than a human: with an LLM, I can backtrack, edit my past words, and get a different answer. One cannot generally do that with a human, as it's a rare trait indeed for someone to be able to hear "you misunderstood, not x, y" and actually take that on board. Even on this site, I've been in threads of:
--
>>>>>>> x
>>>>>> y is dumb, you're dumb
>>>>> sorry you misunderstood, I'm not saying y, I'm saying x
>>>> I just told you y is dumb, you're dumb
>>> no, you're still not getting me, I'm not saying y, I'm saying x
>> stfu noob, I'm a professional: y is dumb, you're dumb
> I'd agree y is dumb, but I'm saying x
stop moving the goalposts!
--
(In this case, x was "it seems implausible that the Saudi Public Investment Fund would really last forever, forever is a long time and there's a always going to be a temptation for the people in charge to extract short term value to the detriment of long term potential", and y was (I think from context but I couldn't read their mind) "they're making bad decisions and will go under imminently".
> I asked chatGPT to program me a complex GUI in Python about selecting multiple time segments out of an overall interval using sliders. It gave me working python code.
Cute. I used it to write a web GUI for its own API, so I can have GPT-4 access as PAYG without a monthly subscription: https://github.com/BenWheatley/YetAnotherChatUI
Sounds like your problems are rather short-term and low-stakes, I've presented it several questions that it cannot answer and tells me so.
While there is certainly untapped potential with this models, what we've seeing is unquestionably market driven hype.
I know very, very few people who are working closely with these models who buy the hype.
I do use GTP* regularly for language learning, or making python templates to extend off of. Is it useful, kinda, not essential imo, but it is not worth as much as it is hyped.
It is possible that "AI" might not be as world-changing as its proponents are claiming. However, if it is free and open-source, and non-commercial, it may still persist and remain useful, regardless of whether it becomes popular or not.
I suspect even if a power ball lottery winner takes a philosophical position it would be taken far more seriously than if they had never won the lottery.
At some point you start to believe your own bullshit when all of society signals are telling you what a genius you are.