Hacker News Reader

Comment by hbarka | Hacker News Reader

This couldn’t be a more definitive comparison of what ChatGPT isn’t good at. WolframAlpha was released in 2009 but why isn’t it getting the same sexy accolades and valuation as ChatGPT?

View on HN

Because most people have no clue it exists. I became aware of it in university and use it for all kinds of queries all the time. There's things made for Google and things made for WA. And now there's things made for ChatGPT. But lots of techies hate WA and so there's not much word of mouth for it outside of academia.

Also most people want funny stuff, like writing jokes or songs, they don't actually want useful information, so chatgpt gives them that. If the internet didn't exist everyone would think it'd be used for research and learning primarily, and mostly it's used for porn and cat videos. It's just what people enjoy.

> But lots of techies hate WA

How could any "techie" possibly hate Wolfram Alpha?

I'm not sure but if you read HN threads they all complain that he is full of himself. I think he built a cool product / company and don't know him so I don't pass judgment myself.

WolframAlpha has a niche use case and rather steep requirements for learning how it wants things to be asked. Also there's a ton of stuff it just can't parse in any format. It can tell you how long to cook an egg (I think) [edit: just checked, it tells you how long to cook ham instead...] but can't tell you any kind of egg recipe.

IMO what you're reading here is just lengthy advertisement for WolframAlpha, written by the guy who created WolframAlpha and named it after himself. The simple queries he's suggesting that WA should be used for aren't even really WA's use case (at least, I've never met anyone who'd bother going to WA to ask it about distance between two cities; Google does this just fine). His argument could be summed up as saying that ChatGPT should just use any legacy search engine to find numerical results and encase the answer in natural language.

Because it's something completely different, which the article also points out:

> Wolfram|Alpha does something very different from ChatGPT, in a very different way.

Imagine an article doing the opposite, showing examples where ChatGPT beats Wolfram|Alpha. It obviously wouldn't even be a comparison worth publishing.

WA beats ChatGPT in a small subset of prompts but those are not what brings the current hype.

Because it only understands language to the extent needed to do math, and that's the easy part.

Merge WolframAlpha and ChatGPT and give it real-time access to the Web, and then things will get interesting.

it's a neat system, but it's basically a fragile natural-language wrapper on very precise calculations. i actually think if you trained chatgpt to spit out the internal wolfram language it could be awesome. (maybe he talks about that in the blog post; i got bored about 20% of the way thru.)

He actually does mention how you could use ChatGPT in tandem with Wolphram|Alpha to produce results that are actually correct, which I think is a very neat idea.

Reply on news.ycombinator.com