As everyone who isn't insane can see, recent AI breakthroughs are very exciting and useful, but also still very limited in what they can accomplish. We are at the beginning of a huge wave of tech coming in as LLMs are straightforwardly deployed for all kinds of things which they're already good at but not yet used for, but nowhere near the dystopian future predicted by the movie Terminator.
There's nothing in the laws of physics which says that we can't make things which outperform artisanal humans at every cognitive task and it seems inevitable that we'll be able to do that some day but when they day may come is still very unclear. We just had a big step up AI but it isn't part of a Moore's Law type of predictable increase, it was a single jump. When jumps will happen in the future is anybody's guess. Maybe we'll have AGI in 10 years, maybe we'll have it in 100 years, there's no way to tell.
(Yes I know GPT has been improving a lot, but I've played with it and can tell you that the jump from GPT3 to GPT4 was nowhere near as big as the jump from GPT2 to GPT3. It's still aggregating things others have said, just doing it with a much larger data set and more memory. This is not to say that GPT4 isn't super impressive, just that there's still a long way to go and new techniques will be necessary to get there.)
Current freakouts over AI safety are counterproductive. In the past there were entities set up to worry about AI safety without actually building AI, and they accomplished diddly. We now have AI projects run by people who care a lot about safety which are both making useful technology and doing it in a way which is very cognizant of safety. Many of the people working at these orgs previously worked at the aforementioned useless ones and decided it was better to be productive.
Different jobs will be affected by LLMs in interesting ways. As is usually the case new technology tends to increase productivity rather than getting rid of jobs completely. Legal briefs for example are very repetitive and a vast corpus of them is by law made public and not copyright encumbered so LLMs can probably do a good job of writing most of them with only a little bit of human guidance and modification.
Management consultants should in principle be worried, because their work output is already a similar level of accuracy and quality to what LLMs spit out, but their lack of useful output never stopped anyone from wasting money on them before.
@solendprotocol This is antithesis of DeFi. If your protocol canβt survive a bull market as it stands then let it burn. Small fires like Solend burning will save the overall crypto market. Stealing from one user to save your so called DeFi exchange is the same thing as big banks being bailed out