[flagged]
It seems that should be able to replace 1st level support for a bank or something like that but those industries move very slowly so even that could be years away still.
Klarna is similar to a bank and they tried replacing 1st level support with a chatbot. Spoiler: it didn't work for them, so I wouldn't keep my hopes so high for banks.
Try disproving your own theory. Can you craft a problem that requires reasoning that an LLM can solve?
You should feel threatened. Maybe they missed this time, but they're shooting at you, and will keep shooting at you.
Not even joking here. If that’s the bar for “reasoning” AI has already beat it.
But that being said, I would say the trendline is the same. LLMs were just an exponential and existential jump in progress that dwarfed anything in the last 15 years. The inevitable slow down after the blip just makes you feel as if it's become much shallower. But in reality, it's the same, and we'll have more and more break throughs.
Obviously the trendline points to the most probable outcome. If you HAVE to bet, bet on probability. It's the most reasonable answer.
Probability can help predict things with enough instances of an event — statistics are only a small part of how we establish a probability of a future event, and a core ingredient of even that small part is having a sufficiently representative sample (which "trendlines" usually don't).
Eg. you can get statistics for all the winning lottery tickets from the last 5 years — that won't help you predict the next one.
Stop confusing the two.
No such pattern exists for the lottery.
Stop being stupid.
I mean, people died to get two days off a week and only work eight hours in the US. That isn’t guaranteed for a wide swath of the world population. And the idea of working only four days a week is crazy radical talk.
This assumes constant progress. I remember during the early days of self driving cars we were told the same: look how far we’ve come already! We’ll all be in self driving cars within 5 years. And yet, here we are.
I have no doubt we will see a ton of progress. I also have no doubt AI will take over a lot of the tedious, basic coding work people do. But replacing coders entirely? I remain skeptical.
It assumes the trendline, as it's constant. Any other assumption is lower probability.
>. I remember during the early days of self driving cars we were told the same: look how far we’ve come already! We’ll all be in self driving cars within 5 years. And yet, here we are.
5 years is a little speculative. 10 is not. At least for me, whenever I call a taxi, it's always an AI taxi. I no longer have human drivers drive me anywhere. So... that's another trendline for you to follow.
And possibly never might.
Sometimes, there is a cap to the capability, and maybe we've already reached it. Yes, there will still be small, incremental improvements. Yes, we can change the "rules" to make new tools more valuable (eg. introduce communicating roads and vehicles to make it easier for cars to be self-driving; or introduce some forms of orthogonal resiliency into software that allows us to live with the slop from LLMs).
And yes, possibly you are right too, but there is nothing definitively confirming that.
I mean, there's a reason "past performance is not a guarantee of future returns" is a well-known idiom.
Try that with the stock market and let us know how it goes (spoiler: many have and they are not rich).
Stocks are the worst example. Trendlines for AI has been pretty reliable for the last decade and we have a North Star that proves we haven’t hit the limit. If a human brain can exist then such intelligence artificial or not can also exist.
Can we achieve it within a practical timeframe of human civilization of, say, 10 thousand years? Or within a single lifetime (some who've worked on it early are now dead, fwiw).
Stocks are obviously a pretty bad example for your claim, because they are a clear counterpoint (in formal logic, which you appealed to, that's enough to call it a proof that trendlines do not ensure any future outcome).
And feedback loop? LLMs have gotten where they are by simply providing current state of the art at no cost to millions of people, to ensure that investment feedback loop by losing gobs of money yet keep investing. We've seen that with VC-funded companies previously too.
The human brain is on earth and exists.
>Stocks are obviously a pretty bad example for your claim, because they are a clear counterpoint (in formal logic, which you appealed to, that's enough to call it a proof that trendlines do not ensure any future outcome).
No they aren't I told you stocks are different. There are many trendlines in the world WHERE scientists use to predict things. So the fact that it's used as a statistical model makes your point completely false. Stocks have one of the most volatile movements.
>And feedback loop? LLMs have gotten where they are by simply providing current state of the art at no cost to millions of people, to ensure that investment feedback loop by losing gobs of money yet keep investing. We've seen that with VC-funded companies previously too.
>And feedback loop? LLMs have gotten where they are by simply providing current state of the art at no cost to millions of people, to ensure that investment feedback loop by losing gobs of money yet keep investing. We've seen that with VC-funded companies previously too.
That's not a feedback loop. People aren't looking at the valuation of the company and trying to bet low or high based off of other peoples attempts to do the same thing.
In stocks people literally LOOK at trendlines and bet based off of that. They both influence the trendline and they take the trendline as input into their own trading algorithm. AI is not doing this at all.
"Trendlines", before being trusted by "scientists to predict things", are carefully examined and only some of them have any "probability" attached to them. Statistics is not probability (which speaks of chance for something to happen), but instead observation of past patterns. Extrapolation into the future is applied very carefully and with sufficient backing.
You also seem to suggest that OpenAI releasing ChatGPT to the public at a huge financial loss has not led to huge investment in generative AI all over the industry and all over the world since there was no "feedback loop" — my point was that this has provided a huge push by getting many smart people working on it after ChatGPT became available to everyone, thus leading to a continuing upward "trendline" in LLM development: it's as self-reinforcing as the example you give with stocks. In statistics, one aberrant huge spike like this one is usually smoothed over a longer time period: the hustlers usually look for the steepest part of the curve and promote that instead.
Another human brain can be as much as 2 feet away from you. You're it. The core of the sun you can't even touch it and it involves unimaginable forces and energy levels never generated on earth. If you can't see the consistency of my logic, then you're illogical. Plain and simple.
>"Trendlines", before being trusted by "scientists to predict things", are carefully examined and only some of them have any "probability" attached to them. Statistics is not probability (which speaks of chance for something to happen), but instead observation of past patterns. Extrapolation into the future is applied very carefully and with sufficient backing.
They have probability. Past patterns are more likely to happen in the future. Extrapolation given what information we have is the best bet. It's not a sure fire bet but it's the best one we have. Your bet of zero information, random gut feelings, anecdotal facts with zero quantification is utter bullshit.
>You also seem to suggest that OpenAI releasing ChatGPT to the public at a huge financial loss has not led to huge investment in generative AI all over the industry and all over the world since there was no "feedback loop" — my point was that this has provided a huge push by getting many smart people working on it after ChatGPT became available to everyone, thus leading to a continuing upward "trendline" in LLM development: it's as self-reinforcing as the example you give with stocks. In statistics, one aberrant huge spike like this one is usually smoothed over a longer time period: the hustlers usually look for the steepest part of the curve and promote that instead.
Did I suggest that? How about no i didn't suggest that. You suggested that. There is no feedback loop genius. AI is tech nobody has ever seen before and they invested into it because of how obviously exciting it is. There is no feedback loop of buying at the peek and selling at the dip by observing a trendline. You're just arbitrarily trying to bend the whole situation to fit the analogy and it only makes your argument worse.
I'm done with you. We can agree to disagree but I'm sorry to say anyone seeing your argument will just laugh. Good day sir.