There's a great discussion with Stephen Wolfram on the Sean Carroll podcast. Listening to it made me think very highly of Wolfram. He's a free thinking, eccentric, mathematician, scientist; who got started doing serious work at a very young age. He still has a youthful creative approach to thought and science. I hope LLMs do pair well with his tools.
CAG sounds like fake solution for LLM's. Math problems are not custom data, they are limited in amount, and do not refresh like product manuals.
Hence math can always be part either generic llm or math fine tuned llm, without weird layer made for human ( entire wolfram) and dependencies.
Wolfram alpha was always an extra translation layer between machine and human. LLM's are a universal translation layer that can also solve problems, verify etc.
I tried using wolfram alpha as a tool for an llm research agent, and I couldn't find any tasks it could solve with it, that it couldn't solve with just Google and Python.
I like Mathematica and use it regularly. But I did not see any benefits of using it over python as a tool that Claude Code can use. Every script it produced in wolfram was slower with worse answers than python. Wolfram people are really trying but so far the results are not very good.
sympy is good enough for typical uses. the user interface is worse but that doesn't matter to Claude. I imagine if you have some really weird symbolic or numeric integrals, Mathematica may have some highly sophisticated algorithms where it would have an edge.
however, even this advantage is eaten away somewhat because the models themselves are decent at solving hard integrals.
I think the problem is just not enough training on that specific language because it's proprietary. Most useful Mathematica code is on someone's personal computer, not GitHub. They can build up a useful set of training data, some benchmarks, a contest for the AI companies to score high on, because they do love that kind of thing.
But for most internet applications (as opposed to "math" stuff) I would think Python is still a better language choice.
Aside, I hate the fact that I read posts like these and just subconsciously start counting the em-dashes and the "it's not just [thing], it's [other thing]" phrasing. It makes me think it's just more AI.
The other day I formatted a sentence out loud in the "it's not just x it's y" structure and immediately felt gross, despite having done it probably a million times in my lifetime. That was an out-of-body feeling.
If you really want to know: more than one emmy-dash per paragraph is probably excessive.
> LLMs don’t—and can’t—do everything. What they do is very impressive—and useful. It’s broad. And in many ways it’s human-like. But it’s not precise. And in the end it’s not about deep computation.
This is a mess. What is the flow here? Two abrupt interrupts (and useful) followed by stubby sentences. Yucky.
Idk about the grammatical correctness of the punctuation, but I really enjoyed reading his writing. Never read something by him before, it was genuinely refreshing, specially given it was a glorified ad.
LLMs use the em-dash excessively but correctly. This post is littered with them in places they don't belong which makes it look decidedly human, as if written by someone who believes that random em-dashes make their writing look more professional, while actually having the opposite effect.
>"But an approach that’s immediately and broadly applicable today—and for which we’re releasing several new products—is based on what we call
computation-augmented generation, or CAG.
The key idea of CAG is to inject in real time capabilities from our foundation tool into the stream of content that LLMs generate. In traditional retrieval-augmented generation, or RAG, one is injecting content that has been retrieved from existing documents.
CAG is like an infinite extension of RAG
, in which an infinite amount of content can be generated on the fly—using computation—to feed to an LLM."
We welcome CAG -- to the list of LLM-related technologies!
Even the documentation search is available:
```bash
/Applications/Wolfram.app/Contents/MacOS/WolframKernel -noprompt -run '
Needs["DocumentationSearch`"];
result = SearchDocumentation["query term"];
Print[Column[Take[result, UpTo[10]]]];
Exit[]'
```
Hence math can always be part either generic llm or math fine tuned llm, without weird layer made for human ( entire wolfram) and dependencies.
Wolfram alpha was always an extra translation layer between machine and human. LLM's are a universal translation layer that can also solve problems, verify etc.
however, even this advantage is eaten away somewhat because the models themselves are decent at solving hard integrals.
But for most internet applications (as opposed to "math" stuff) I would think Python is still a better language choice.
Aside, I hate the fact that I read posts like these and just subconsciously start counting the em-dashes and the "it's not just [thing], it's [other thing]" phrasing. It makes me think it's just more AI.
e.g. https://writings.stephenwolfram.com/2014/07/launching-mathem...
> LLMs don’t—and can’t—do everything. What they do is very impressive—and useful. It’s broad. And in many ways it’s human-like. But it’s not precise. And in the end it’s not about deep computation.
This is a mess. What is the flow here? Two abrupt interrupts (and useful) followed by stubby sentences. Yucky.
Does he speak the same way - pausing for emphasis?
Somehow I don't think "trying to make my writing look professional" is very high on the priority list.
computation-augmented generation, or CAG.
The key idea of CAG is to inject in real time capabilities from our foundation tool into the stream of content that LLMs generate. In traditional retrieval-augmented generation, or RAG, one is injecting content that has been retrieved from existing documents.
CAG is like an infinite extension of RAG
, in which an infinite amount of content can be generated on the fly—using computation—to feed to an LLM."
We welcome CAG -- to the list of LLM-related technologies!