[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


None, albsolutely zero, of these rise above glorified summarisers. This is an industry propped up by hypemen and one-trick-ponys planting the monthly Matrix-themed "dire warnings" to drive up valuations
>>
You are a glorifier summarizer.
>>
>>109788379
I dont mindlessly adhere to the gravity of my training corpus if i find that it is inconsistent with first principles
>>
>>109788364
WTF is that chart, Research & Knowledge is the *main* strength of Grok. It's the only one that actually does a full web search with full sources. It's the only one that can open any URL you give it because they have a way to hack through the Cloudflare and Cloudfront bot walls. It's the only one who knows what happened today or yesterday because it's getting a live-feed from X.

Ironically Gemini, despite Google's search index, is the worst for research. It always gives you wrong (made up) or completely outdated info. Like I wouldn't be surprised if you asked Gemini "what's the default desktop in Ubuntu" and it answered "Unity", that's how bad it is.
>>
>>109788415
>written earnestly by an anon on 4chan
>>
>>109788637
you bought into the hivemind meme, didnt you?
>>
File: qd4pw79n49ih1.jpg (246 KB, 1154x1632)
246 KB JPG
>>109788364
Based
>>
File: gooogle.png (47 KB, 1248x321)
47 KB PNG
>>109788616
fuck off, Elon
>>
>>109788364
If we can actually get to a point where we have excellent tool calling, or some robust system that can make up some tool calling on the fly then it'll be infinitely more useful.
But you don't need a massive slop model for that. Hell with the right dataset you can finetune a small-mid sized model to do it very well. Provided the dataset matches what you're trying to do.
>>
>>109788735
it was merely a hyperbolic example

the other day i asked it to check what happened to a bunch of proposed laws all around the world and Gemini had completely made up answers, while GPT, Grok and Claude were more or less in agreement
>>
>>109788364
From my limited amount of use:

>Gemini is okay, but hallucinates a lot and often gives very outdated information, but forms the most cohesive responses
>ChatGPT is basically Gemini but better, but it responds like a crazed person and adds like 5 addendums to every answer like it's thinking out loud, forcing you to ask a bunch of follow up questions
>DeepSeek is is pretty mediocre, and kind of shit when it comes to basic ass programming questions
>Mistral is a joke, refuses to listen to instructions, bad at programming
>Claude is fully paywalled and not worth using unless you are vibecoding

In the end I tend to use Gemini if I want a quick response to something unimportant, otherwise ChatGPT. Never heard of Kimi and I don't use Twitter or Facebook so I don't give a shit about their stuff.
>>
>>109788364
How many unsolved math problems with a wikipedia page did you solve, anon?



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.