throwaway4aday

11 ms·

61.
by throwaway4aday 2y ago
Forgot to share this link as well, not sure if you're aware of it but it's a great write up on fine tuning small local models on specific APIs and seems like it would be a perfect fit for your project. https://bair.berk
62.
by throwaway4aday 2y ago
Love it! Voice interaction is a great modality for UI. A lot of people have a bad taste left over from early attempts but I expect to see a lot of progress made now that STT and natural language understanding is so much better. The biggest
63.
by throwaway4aday 2y ago
Using Whisper as the voice interface, an LLM to understand the prompt and issue function call commands and an image upscaler you could build this in a weekend. Would it be useful? Not especially by itself but I think there is a lot of promi
64.
by throwaway4aday 2y ago
That's called Websim my friend and it's a heck of a lot of fun https://websim.ai This is what it generated for one of the non-existent articles https://websim.ai/c/P32MdBI15Ytxwwlth
65.
by throwaway4aday 2y ago
It could be done and it might actually be useful. You could fine tune one of OpenAI's models on the material owned by the company and use RAG to create bespoke summaries or quotations to fulfill the user's request but if you'
66.
by throwaway4aday 2y ago
If their multi-modal model works similarly to any of the existing voice models then they would only need a fairly small number of samples. Current voice models only require you to prefix a voice sample in order to mimic it. They also still
67.
by throwaway4aday 2y ago
I'm expecting much less drama. Maybe AI startups will fail at a higher rate than regular startups but there will still be success stories. Are we conveniently forgetting that the vast majority of businesses fail? There are still crypto
68.
by throwaway4aday 2y ago
Holy buried lede Batman! Right at the end. > Abliteration is not limited to removing alignment and should be seen as a form of fine-tuning without retraining. Indeed, it can creatively be applied to other goals, like FailSpy's Mopey
69.
by throwaway4aday 2y ago
If they published it freely then they wouldn't get paid. I thought everyone was mad about how there aren't enough employment opportunities for PhDs? Training LLMs on your area of expertise is surely preferable to working a dull of
70.
by throwaway4aday 2y ago
> Usage data: ChatGPT is said to generate on the order of 10 billion tokens of data per day – even before they opened their more compelling GPT-4o model to free users. > Common Crawl (filtered) 410 billion tokens 60% of GPT-3 training
71.
by throwaway4aday 2y ago
Solving that will be a much bigger deal but it's at odds with producing a highly accurate emulation of human thought and language. Language models can serve as tools to understand and experiment with logic formulated as natural languag
72.
by throwaway4aday 2y ago
Why not? Nuclear is far safer than a coal or gas plant and that's using the older model reactors as a stats source. Newer small reactor designs are even safer. Anxiety and fear of nuclear power is a purely media and activist driven phe
73.
by throwaway4aday 2y ago
Why fight an uphill battle for reduction in manufacturing when you can get rich by being the first to offer cost competitive on-site carbon free power production? Forget marketing rooftop solar to households, you should be selling micro-nuc
74.
by throwaway4aday 2y ago
Listen to the side by side comparisons. Sky has a deeper voice overall, in the gpt4o demo Sky displays a wider pitch range because the omni model is capable of emotional intonation. Her voice slides quite a bit while emoting but notably doe
75.
by throwaway4aday 2y ago
I like a lot of these ideas. Some of this is built in to the ChatGPT desktop app. Other parts of it could be glued together from existing tools. Others are still beyond the capability of LLMs. Lots of people thinking along the same lines an
76.
by throwaway4aday 2y ago
That's the ChatGPT desktop app
77.
by throwaway4aday 2y ago
It could be misused and leak vital data but so can your browser history and the data that ISPs and social media and search companies have on you. They say all processing will be done locally but you'd have to be a fool to trust that. I
78.
by throwaway4aday 2y ago
The training process is not purely mechanical. The amount of time that goes into selecting, cleaning, reformatting and otherwise preparing the training data is significant not to mention all of the other work involved in the actual training
79.
by throwaway4aday 2y ago
I think it's as clear as it can be, they go into much more detail and provide examples in their bullet points, here are some highlights: Our model learns from previous suggestions and whether or not a user joins the channel we recommen
80.
by throwaway4aday 2y ago
Considering how much illogical and mistaken thought and messy, imprecise language goes into achieving logical reasoning I honestly don't think there will ever be such a thing as "logical AI" if by that you mean something whic
81.
by throwaway4aday 2y ago
My uninformed question about this is why can't we make the VRAM on GPUs expandable? I know that you need to avoid having the data traverse some kind of bus that trades overhead for wide compatibility like PCIe but if you only want to u
82.
by throwaway4aday 2y ago
Speaking of physics, we should borrow the quote "Shut up and calculate" to describe the situation: it works so use it now and worry about the explanations later.
83.
by throwaway4aday 2y ago
Honestly, I don't care if other people are creating similar things to what I am. Actually, I prefer if there are more people working on the same things because it means there are other people that I can talk to about those things and c
84.
by throwaway4aday 2y ago
You get to use unreleased models for free and compare them to others, you are gaining some benefit from this. Why be annoyed about it? They aren't making you do anything, you're choosing to play around with these models in a &quot
85.
by throwaway4aday 2y ago
I don't know about the rest of the developers in the world but my dream come true would be a computer that can write all the code for me. I have piles of notebooks and files absolutely stuffed with ideas I'd like to try out but be
86.
by throwaway4aday 2y ago
I don't see a flattening. I see a lot of other groups catching up to OpenAI and some even slightly surpassing them like Claude 3 Opus. I'm very interested in how Llama 3 400B turns out but my conservative prediction (backed by Met
87.
by throwaway4aday 2y ago
can you select a context length that fits in your GPU though? I suppose even a 128k model would be more than enough for almost everyone running these models on their own hardware.
88.
by throwaway4aday 2y ago
a simple solution is to put a small submersible fountain pump in or even one of those floating solar powered ones. mosquitos prefer standing water so if you agitate it enough it will cut down on their breeding. please do make every effort t
89.
by throwaway4aday 2y ago
This assumes we've already achieved, or can achieve through efficiency gains that can be developed faster than generating capacity can be built, an optimal amount and cost for electricity. I'd like to see anyone argue that. If ele
90.
by throwaway4aday 2y ago
China has successfully built many reactors in as little as 5 years[0] though construction times vary for others. Historically, other countries have been able to rapidly build out nuclear capacity with similar or better construction times. W

More ›