PianothShaveck
@pianoth
Followers
53
Following
4K
Statuses
1K
Pianist and composer.
Salerno, Italy
Joined December 2019
You can follow me on spotify: and you can also subscribe on my Youtube channel:
0
1
2
I cancelled the OpenAI subscription. I see no point for most people to get the $20 plan; I certainly see no point anymore for myself. If I needed to specifically use an OpenAI LLM, an API request every once in a while is more than sufficient. I'm not buying the hype, personally.
0
0
1
@kimmonismus If you do any tech job, maybe, sure, if they solve the random hallucination problem (which we're pretty much nowhere near solving). But, aside from those, we're nowhere near having an AI that could replace a music teacher. Or an art teacher. And yes, those jobs are still needed.
1
0
1
0
415
0
On the other hand, having a list of hundreds of mechanical steps and an insanely complex construction doesn't help at all. This is what AG2 does, and this is why it's actually not that impressive nor useful.
0
0
1
Ok tried out Gemini Pro: not impressed. Feels about on par with 4o, there's a slight improvement over the previous exp-1206, but it is still messy in the way it writes step by step, and it's obviously no where near as accurate or reliable as a thinking model like o1 or o3-mini
0
0
2
Excited to try out the new gemini models, I hope they'll also make a Pro Thinking variant
0
0
1
@yunjinzone Whoa Lan Yan top 3? My top 4 is probably Yun Jin ≈ Chiori > Furina > probably Yoimiya but I don't have a stable fourth
1
0
1
I tested o3-mini-high on a currently unsolved problem by all LLMs. Its solution was incorrect, but it really seemed correct. It took me 20 minutes to understand why it wasn't. This is concerning, because this means o3 is able to write in a way that is extremely persuasive.
0
0
2
