"It's autocomplete" is not a helpful analogy to understand LLMs. A LLM is more like a database that lets query information in natural language. You can query both knowledge, and "patterns" (associative programs seen in the training data, that can be applied to new inputs).
@fchollet
-
Replication and AGI: Equal Distance from Current Reality
By
–
We are as far today from arbitrary matter replication and grey goo as from AGI.
-
No Clear Path from Current Technology to Sci-Fi Evolution
By
–
The thing is, in both cases there is *no path* from the current technology (e.g. fused deposition modeling, or deep learning) to its hypothetical sci-fi evolution. What you're working on sounds superficially similar, and may well change the world. But it's a different thing.
-
3D Printing Arguments More Convincing Than AGI Threat Narratives
By
–
The 3D printing company would actually have a much more convincing argument here, because the mechanism through which matter replication brings infinite prosperity is easy to articulate, and so is the dire threat represented by grey goo. Whereas for AGI it is all left implicit
-
Mundane Technology Hyped as Infinite Returns and Existential Threat
By
–
In both cases, you have a useful but fairly mundane technology, and you pretend it will soon bring *infinite* returns (on the basis that it is depicted as such in sci-fi), while representing a major existential threat (same reasons).
-
AI Companies and AGI: An Analogy with 3D Printing
By
–
When companies that train deep learning models talk about AGI, it's as if a 3D printing company talked about how the next generation of the technology was going to bring universal abundance by enabling arbitrary matter replication — if we can avoid the grey goo scenario
-
LLMs as a Solution to Poor Language-Based User Experience
By
–
You could say that LLMs are a general way to fix bad UX in human systems that use language as their interface. Of course, in a better world, this would be fixed via user-friendly online forms instead.
-
LLMs for Document Boilerplate Generation: Solving User-Hostile System Interfaces
By
–
One of the most valuable applications of LLMs: generate document boilerplate for you (e.g. an official letter). It's a UX improvement: the reason this is a problem that needs solving in the first place is that so many human systems have user-hostile interfaces.
-
ML Deployment and Productivity Stagnation Over the Past Decade
By
–
For all the talk about ML destroying jobs, productivity had been improving fast long before we used any ML, and over the past 10 years we went from "very little ML" to to "ML deployed everywhere" and productivity stayed stagnant