3 ms·

It's still largely unclear how AI is going to affect the job of software engineers both in the shorter term, pre-AGI, and after that. However, I'm pretty sure

It's still largely unclear how AI is going to affect the job of software engineers both in the shorter term, pre-AGI, and after that. However, I'm pretty sure that the job of SWE isn't about to disappear, since even when human-level AGI appears, the process of software development isn't about to get any easier than when it was done by actual humans. I don't see non-technical middle managers able to take over as systems architects and lead developers. It'll just mean that the day-to-day of those charged with developing software per PHB's requirements changes. That said, I do feel fortunate to have been from the generation where software was developed entirely by hand, having started in the era of 8-bitters (NASCOM-1 in 1978) and only just in the last year transitioning (against my preference) to a role of overseeing offshore developers rather than being hands-on. Still, I think that those who enjoy the challenge and fun of conjuring up dreams out of software will still find plenty to enjoy in the future, although it will be different, just as it has already changed a lot from the time when we were writing everything in assembler to today arguing about whether Rust is better than C++.

zwnow 1y ago
After AGI? AGI is not a reality anyone of us will witness. Its a false promise in the current LLM market and we are as far away from it as we always have been... just listen to actual experts on the topic that dont profit from pushing a certain narrative.
- HarHarVeryFunny 1y ago
  Demis Hassabis is estimating 5-10 years to AGI, and is about the only person who's opinion on this I think is worth listening to. His/their (DeepMind) plan is to retain a pretrained LLM but add something like search around it (cf AlphaGo or AlphaFold). He does say a couple of "Transformer-level" major breakthoughs may be needed to get there, and is still only 50% confident because of AGI-requirements he may have missed. Hassabis has the Google DeepMind team of 5000+, very many of who are PhD's, working for him, so a lot can be done in 5-10 years.
  - zwnow 1y ago
    LLMs can not reason. We'd need a very very different technology than LLMs to get there. Not to mention new "models" are basically old model agents with access to some more tools. Training actual new models is currently not cost efficient. Also LLMs plateaued, idk how Mr. Hassabis got to that conclusion but I call bs. A "couple" major breakthroughs can take 1000s of years.
    - HarHarVeryFunny 1y ago
      An LLM by itself can only regurgitate reasoning and/or reasoning steps from the training set, but I think that adding search on top of it gets you much closer. You're basically talking about the possibility of searching through all possible sequences of reasoning steps that an LLM could generate, and picking the sequence that actually works. DeepMind did it for AlphaGo and AlphaFold - I would not bet against them. When Hassabis says it'll take 5000+ people 5-10 years and maybe a couple of Transformer-level breakthroughs, then he's clearly not talking about just adding search, and everyone realizes that things like continual/incremental learning are also required. I'd guess that Hassabis has a pretty good idea of what's missing from LLMs and needs to be added to get to AGI. I really would not dismiss Hassabis. He is a lot smarter than you or I, and has won a Nobel Prize for his application of AI to science, as well as stunning the machine learning community with AlphaGo (most had thought beating Go would take another 10 years). I think there are better long term approaches to super-human AI than building upon LLMs, but just as a cog is not a car, there is no reason it can't be a part of a car.

zwnow 1y ago

After AGI? AGI is not a reality anyone of us will witness. Its a false promise in the current LLM market and we are as far away from it as we always have been... just listen to actual experts on the topic that dont profit from pushing a certain narrative.

HarHarVeryFunny 1y ago
Demis Hassabis is estimating 5-10 years to AGI, and is about the only person who's opinion on this I think is worth listening to. His/their (DeepMind) plan is to retain a pretrained LLM but add something like search around it (cf AlphaGo or AlphaFold). He does say a couple of "Transformer-level" major breakthoughs may be needed to get there, and is still only 50% confident because of AGI-requirements he may have missed. Hassabis has the Google DeepMind team of 5000+, very many of who are PhD's, working for him, so a lot can be done in 5-10 years.
- zwnow 1y ago
  LLMs can not reason. We'd need a very very different technology than LLMs to get there. Not to mention new "models" are basically old model agents with access to some more tools. Training actual new models is currently not cost efficient. Also LLMs plateaued, idk how Mr. Hassabis got to that conclusion but I call bs. A "couple" major breakthroughs can take 1000s of years.
  - HarHarVeryFunny 1y ago
    An LLM by itself can only regurgitate reasoning and/or reasoning steps from the training set, but I think that adding search on top of it gets you much closer. You're basically talking about the possibility of searching through all possible sequences of reasoning steps that an LLM could generate, and picking the sequence that actually works. DeepMind did it for AlphaGo and AlphaFold - I would not bet against them. When Hassabis says it'll take 5000+ people 5-10 years and maybe a couple of Transformer-level breakthroughs, then he's clearly not talking about just adding search, and everyone realizes that things like continual/incremental learning are also required. I'd guess that Hassabis has a pretty good idea of what's missing from LLMs and needs to be added to get to AGI. I really would not dismiss Hassabis. He is a lot smarter than you or I, and has won a Nobel Prize for his application of AI to science, as well as stunning the machine learning community with AlphaGo (most had thought beating Go would take another 10 years). I think there are better long term approaches to super-human AI than building upon LLMs, but just as a cog is not a car, there is no reason it can't be a part of a car.