LARRY ELLISON: AI IS RAPIDLY COMMODITIZING BECAUSE MOST MODELS ARE TRAINED ON THE SAME PUBLIC INTERNET DATA.
THE REAL COMPETITIVE EDGE ISN’T THE MODEL ANYMORE — IT’S ACCESS TO EXCLUSIVE, PROPRIETARY DATASETS.
THAT MAY BE THE ONLY MOAT LEFT.
@naksterz It's 100% probabilistic. Mapping out relationships between words and texts and predicting, based on context, which is most likely to fit in that context.
1/ LLMs are next token prediction systems. I think people overlook their probabilistic nature.
The conditional probability of being useful increases as context is provided.
IMO, this is why PLTRs software is receiving such great reviews. PLTR helps provide the context to LLMs
@tandoorinightts Better to start looking and leave before you get fired. Trust me on that one. If it's that bad after 1.5 months, it's going to get worse and you will probably end up getting fired or laid off.
@shashankmodi_@Nithya_Shrii If your management style involves managing the fine details, to the point that you are managing away all the decisions on the project, then you are a bad manager.
@taba06794571@PaulSkallas Savate guys don't do well in MMA because there are basically only a few Savate guys. I am friends with a multi-time Savate world champ and he is a monster. He's in his mid 40s and spars hard all the time, beats the crap out of up and coming semi-pro boxers in pure boxing sparring
@PaulSkallas 1. That throw would not knock someone out on concrete.
2. Judo is awesome and works great. However, upper body kimono throws are not as effective against a shirtless and slick opponent.
@DeepValueBagger@No1caresellmore That's an interesting opinion. In my opinion, there are going to be a greater number of tech companies, with fierce competition, rapid iteration and product releases. The end result will not be the dying of software, but insane hyper-growth of new tech and economic expansion.
@DeepValueBagger@No1caresellmore That's a possibility. It's also possible that demand grows because the economic utility of each worker is so much greater. I guess we will find out in a few years. It's hard to say exactly how things will end up in the post-AI world.
@DeepValueBagger@No1caresellmore What is true is that entry level jobs are screwed. They are down by 80-90%. But, it's not true that SWE job demand is trending down. It's actually trending up and, yes, some companies do layoffs. But, it's easier to say "AI" than it is to say "we screwed up and need to cut"
@DeepValueBagger@No1caresellmore I'm not saying that you are wrong about that. But, I don't agree with your analysis of Meta's layoffs. In the paradigm you are describing, SWE is a skill set and knowledge base that is more powerful and has more economic utility. It's counterintuitive, but that's my read
@DeepValueBagger@No1caresellmore I don't agree with that analysis, because, IMO, the data indicates otherwise. We see a lot of these AI-native companies hiring like crazy and the demand for software engineering is actually starting to trend up (although at the mid and senior level). IMO layoffs are unrelated
@DeepValueBagger@No1caresellmore OK, so let's analyze that premise. I'm don't disagree with experts becoming super powered. So, software engineers actually become more valuable, because their economic utility is increased. So, why would you lay them off, if they are more valuable?
@DeepValueBagger@No1caresellmore i.e, does Claude, every single time, produce the absolute best most performant code? I am skeptical about that, especially considering how many of Anthropic tools are known to be buggy on the time of release.
@DeepValueBagger@No1caresellmore I'm not saying Claude can't write C. What I'm saying is that there are varying degrees of outcomes that are possible with respect to some type of super high performance application. When you stand something up in a week end, with vibe coded tools, I'm skeptical about the quality.