Hacker News Reader

Hacker News

Comment by petilon | Hacker News Reader

Full thread

petilon

With Wolfram plugin ChatGPT is going to become a Math genius.

OpenAI is moving fast to make sure their first-mover advantage doesn't go to waste.

View on HN

I guess I'm a bit vindicated from my prediction 40 days ago!

"GPT needs a thalamus to repackage and send the math queries to Wolfram"

https://news.ycombinator.com/item?id=34747990

Stephen Wolfram himself thought that Wolfram could be combined with GPT when ChatGPT was released like 4 months ago. Only due to that they worked together to build the plugin. He also authored the best article I've seen on how ChatGPT and more broadly LLMs work (that has now been turned into a book).

There's a book? Where's the book?

https://a.co/d/9HasuKR

I was drawn to the Wolfram logo blurb as well. It is funny because within days of ChatGPT making waves you had Stephen Wolfram writing 20,000-word blog posts about how LLM's could benefit from a Wolfram-Language/Wolfram Alpha API call to augment their capabilities.

On one hand I'm sure he will love to see people use their paid Wolfram Language server endpoints coupled to OpenAI's latest juggernaut. On the other, I'm sure he's wondering about what things would have looked like if his company would have been focused on this wave of AI from the start...

I'm very excited for GPT to summarize Stephen Wolfram's writing.

This too is one of the most interesting integration to me. Allows for getting logical deduction from an external source (e.g. wolfram alpha), which can be interacted with via the natural language interface. (e.g. https://content.wolfram.com/uploads/sites/43/2023/03/sw03242...)

For those interested the original Stephen Wolfram post:

https://writings.stephenwolfram.com/2023/01/wolframalpha-as-...

And the release post of their plugin:

https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its...

I feel like people with smart AI's would have an advantage in making smart decisions. Probably at this point they discuss business strategy with some version of it.

Why cant Wolfram train a rudimentary chat model in their own search box. it doesn't even need to be very knowledgeable, just know how to map questions to mathematica

more accurately, chatgpt is already quite good at mathematical concepts, it just has difficulty with arithmetic due to tokenization limitations: https://www.lesswrong.com/posts/qy5dF7bQcFjSKaW58/bad-at-ari...

Reply on news.ycombinator.com