ainch

10 ms·

151.
by ainch 8mo ago
Moltbot is an attempt to do that. Would you hire it as a personal secretary and entrust all your personal data to it?
152.
by ainch 8mo ago
As a user that doesn't pay for the subscription, I use their Google Drive integration quite often. I believe they support Dropbox and OneDrive too, no?
153.
by ainch 9mo ago
It's a good paper by Hooker but that specific comparison is shoddy. Llama and Aya were both trained by significantly more competent labs on different datasets to Falcon and BLOOM. The takeaway there is "it doesn't matter if y
154.
by ainch 9mo ago
I'm not going to bat for GPTZero, but I think it's clearly possible to identify some AI-written prose. Scroll through LinkedIn or Twitter replies and there are clear giveaways in tone, phrasing and repeated structures (it's n
155.
by ainch 9mo ago
OpenAI are testing ads in the free tier of ChatGPT, but they state that the actual LLM responses won't include advertising/product placement [0]. [0]: https://openai.com/index/our-approach-to-advertising-and-e
156.
by ainch 9mo ago
I'm not sure Anthropic will become ad-supported - the vast bulk of their revenue is b2b. OpenAI have an enormous non-paying consumer userbase who are draining them of cash, so in their case ads make a lot more sense.
157.
by ainch 10mo ago
As LLMs are productionised/commodified they're incorporating changes which are enthusiast-unfriendly. Small dense models are great for enthusiasts running inference locally, but for parallel batched inference MoE models are much m
158.
by ainch 10mo ago
Perhaps they need more advertising around the correct spelling of his name.
159.
by ainch 11mo ago
Gemini 3.0's cutoff is January. I think you can get away with it if the model has good search/tool use capability.
160.
by ainch 11mo ago
Which other group is that?
161.
by ainch 11mo ago
Of course people don't say it, but there are many cases where reported algorithmic improvements are attributable to poor baseline tuning or shoddy statistical treatment. Tao is exhibiting a lot more epistemic humility than most researc
162.
by ainch 11mo ago
Inferior in what sense? Genie 3 is addressing a fundamentally different problem to a physics sim or procgen: building a good-enough (and broad-enough) model of the real world to train agents that act in the real world. Sims are insufficient
163.
by ainch 11mo ago
That's a good correction, thanks.
164.
by ainch 11mo ago
The reports are definitely bland, but I find them very helpful for discovering sources. For example, if I'm trying to ask an academic question like "has X been done before," sending something to scour the internet and find me
165.
by ainch 11mo ago
Generally you train each expert simultaneously. The benefit of MoEs is that you get cheap inference because you only use the active expert parameters, which constitute a small fraction of the total parameter count. For example Deepseek R1 (
166.
by ainch 11mo ago
That's an interesting idea, it sounds similar to the principles behind low precision models like BitNet (where each weight is +-1 or 0). That said, I know Deepseek use fp32 for their gradient updates even though they use fp8 for infere
167.
by ainch 1y ago
Cutting people at FAIR is a real shame though - great models like DINO and SAM have had massive positive impact - hopefully that work doesn't slow in favour of LLM-only development at MSL.
168.
by ainch 1y ago
I think you're overstating the distinction between ML and generation - plenty of ML methods involve generative models. Even basic linear regression with a squared loss can also be framed as a generative model derived by assuming Gaussi
169.
by ainch 1y ago
There is a lot of overlap between AI and Neuroscience, especially among older researchers. For example Karpathy's PhD supervisor, Fei-Fei Li, researched vision in cat brains before working on computer vision, Demis Hassabis did his PhD
170.
by ainch 1y ago
Their recent paper suggests the active user base is continuing to grow consistently with consistent/growing usage based on how long they've been using the app. https://cdn.openai.com/pdf/a253471f-8260-40c6-a2c
171.
by ainch 1y ago
Gemini 2.5-Pro was great when it released, but o3 and GPT-5 both eclipsed it for me—the tool use/search improvements open up so many use cases that Gemini fails at.
172.
by ainch 1y ago
I doubt this is coming from RLHF - tweets from the lead researcher state that this result flows from a research breakthrough which enables RLVR on less verifiable domains.
173.
by ainch 1y ago
Have you tried the Canvas feature for collaborative writing? Agreed on voice mode - would be great to be able to narrate while doing busywork round the house.
174.
by ainch 1y ago
I don't know what price you'd pay in your hypothetical scenario, but I would say that European universities often take a larger chunk of equity than US unis where spinouts are concerned [1]. Most US universities take between 0-5%
175.
by ainch 1y ago
The technical report does go into a lot of depth about how they use RL, such as the modified GRPO objective they use. As far as the README, I imagine most people active in the field understand the implications of "RL" for a reason