Hacker News Reader

Comment by hcarvalhoalves | Hacker News Reader

That is a pretty grounded, timeless write-up for 1960.

    5.4 The Language Problem

    The basic dissimilarity between human languages and computer languages may be the most serious obstacle to true symbiosis.

LLMs unlock the much sought after capacity to query data in natural language, that is why "AI" is such a hype, besides not being intelligent (in terms of reasoning) at all.

View on HN

Licklider was primarily referring to the need to suppercede punch card languages with ones with an interactive human language interface. Chiefly, McCarthy's Lisp, which he directed the Arpa/IPTOs funding towards, but it would apply to practically every language since - though in terms of interactivity still CL most of all.

That said he also lead the funding of AI, so LLMs are not outside of what he was imagining altogether, for example he wrote about the need for a busy executive to communicate with his computer like he would his secretary, etc. But he also seemed to think that interactive programming environments could accomplish this goal.

Sidenote, something that's forgotten is that Lick was a Lisp person. His archives contain his notes to SICP, and various toy programmes.

The computer works with binary sequences. LLMs do not change this fact.

They don't really have to. Even in Star Trek where you can tell the holodeck to do whatever in LLM fashion, there's also a lower level language that can be used for more fine control. It's always good to have levels of abstraction when you need them, this just adds the very top one that was always missing.

The fact is that they do and it doesn't matter how many abstractions we put on top of the underlying physical reality of digital computation, it will always be digital unless someone figures out how to make analog computation as scalable as digital computation.

Reply on news.ycombinator.com