Linus Torvalds slams AI as '90% marketing and 10% reality'

I’m with him.<p>AI is cool but its mainstream appeal right now is an improved search.<p>All the other forms of generative AI are gimmicks. All classification ai seems to have hit a limit in how useful it can be. Take recognition. It’ll never be 99.999%, and so we can only depend on it so much.

Let&#x27;s focus on LLMs because I think that is the gist of this article. Whilst Linus may not like the BS, as most of us don&#x27;t, I disagree that LLMs are not useful. There are many companies now implementing LLMs, even the EU parliament has implemented Claude and there are a growing list of use cases from banks, pharma, FMCG and others.<p>The article also digresses by stating &quot;companies are struggling to deploy meaningful AI tools for a variety of reasons, including poor infrastructure and a lack of readiness on the parts of both workers and leaders.&quot; That is not because the LLMs are not useful - it is actually fear and inertia. So many companies are afraid to implement LLMs because a) horror stories on helpfullness, harmlessness and honesty of LLMs and b) fear of the time and infrastructure it will take because big IT projects took years and were over budget - but LLMs are different - the EU implemented Claude in less than 6 months on millions of disparate data and systems.

While i somewhat agree, the real sad thing is that 10% is real and usefull, yet it gets overshadowed by that 90% hype based products.<p>Think of how mych money is being thrown at hype products that could have been allocated to the next Cancer detection AI, or AI prosthetic, but instead these things have to take 2nd place behind stuff like the Rabbit R1, or Devin, or the AI washing machine.<p>At least their is a somewhat usefull product this time, unlike with NFT.<p>Besides, i was like 3 when 2008 happened, and it dodnt really happen in my country, so it would be nice to experience what a real recession feels like :v

Slams? I&#x27;d say that&#x27;s being generous, if anything.