jdminhbg

13 ms·

151.
by jdminhbg 2y ago
That analogy doesn't work. `train` is not a deterministic process. Meta has all of the training data and all of the supporting source code and they still won't get the same `weights` if they re-run the process. The weights are the
152.
by jdminhbg 2y ago
Yes, you can finetune Llama: https://llama.meta.com/docs/how-to-guides/fine-tuning/
153.
by jdminhbg 2y ago
I agree that you need training data to build AI from scratch, much like you need lots of really smart developers and a mailing list and servers and stuff to build the Linux kernel from scratch. But it's not like having the training dat
154.
by jdminhbg 2y ago
> In any case, if a program was built by training against a dataset, the whole dataset is part of the source code. I'm not sure why I keep seeing this. What is the equivalent of the training data for something like the Linux kernel?
155.
by jdminhbg 2y ago
> Why do people keep mislabeling this as Open Source? The whole point of calling something Open Source is that the "magic sauce" of how to build something is publicly available, so I could built it myself if I have the means. B
156.
by jdminhbg 2y ago
Summed up in a tweet: https://x.com/JucheMane/status/852949235981012994
157.
by jdminhbg 2y ago
> I'm sure that there is Somebody doing all these things he's talking about, but I kind of have to wonder if it drives him that crazy why is he still listening? It is pretty annoying when sources of music news and reviews you u
158.
The EU Goes Too Far(stratechery.com)
30 points by jdminhbg 2y ago|74 comments
159.
"Bullshit Jobs" Is a Terrible, Curiosity-Killing Concept(thediff.co)
1 points by jdminhbg 2y ago|1 comments
160.
by jdminhbg 2y ago
I think the other one fell off because of the flamewar detector, compare the number of upvotes to the number of comments.
161.
by jdminhbg 2y ago
2 I guess you can argue; they say that's all encrypted and they're sharing the OS with security researchers to audit. 3 I don't think is applicable, OpenAI is not getting screenshots for awareness, they're just getting &
162.
by jdminhbg 2y ago
Ok, that doesn't really work either, because the Apple Intelligence stuff is done on-device.
163.
by jdminhbg 2y ago
I'm pretty sure that SharePlay screensharing is encrypted like FaceTime is, so I don't think you're sharing that stream of screenshots with Apple by using it. Their concern would be that if you allow third parties to use that
164.
by jdminhbg 2y ago
> Apart from the “security concerns” of allowing use of non Apple services I would think that if the EU requires Apple to allow third parties to use the new screen sharing APIs with the same privileges, that would be a security concern.
165.
by jdminhbg 2y ago
> Perhaps you aren't familiar with the term. Graduate Student = someone enrolled in a Masters Program, Post doc, PHD program, etc. I'm familiar, thanks. Why is someone who graduated last year and went into a Master's progr
166.
by jdminhbg 2y ago
> TAs in colleges are graduate students. They aren't undergraduates who can't find a job. So they're... graduates who can't find a job? I'm not sure you're making the point you're trying to make here.
167.
by jdminhbg 2y ago
This is about the entire lifespan of the Starliner program, not just one piece of hardware.
168.
by jdminhbg 2y ago
If I quit a job because they want everyone to return the office, was I fired?
169.
by jdminhbg 2y ago
> They really can't, because a majority of their compensation is in now-unreasonable amounts of OpenAI equity (because it was worth much less when it was granted). If employees decided to walk away, rendering their vested equity wor
170.
by jdminhbg 2y ago
OpenAI is substantially different from other companies in that the capital part is almost totally ephemeral (the Azure credits). Amazon employees can’t just walk across the street and start a new company because they’d be leaving all the wa
171.
by jdminhbg 2y ago
What they didn't realize is that they were going to have to convince the rank-and-file employees of that. Without a communication strategy to do that, they had basically no leverage.
172.
by jdminhbg 2y ago
> On paper the board has the power and it’s their job to monitor and hire/fire the execs. But they’re dependent on the execs for info. It's way worse for the OpenAI board position in this case, because the company is basically
173.
by jdminhbg 2y ago
> I still love being in a ruby environment more than a JS one mostly because it all feels so straightforward. Somehow, this is an ethos/style baked into Ruby, whereas javascript is too spastic to be straightforward. A blog engine is
174.
by jdminhbg 2y ago
> Unfortunately they pivoted to a business model of maximum profitability Based on the article, they still haven't manage to pivot to any profitability.
175.
by jdminhbg 2y ago
Who said it's trademarked? Apple has apps on macOS named "Photos" and "Calendar." There's apps on iOS named "Camera" and "Notes." It's not at all crazy to have one called "TV.&quot
176.
by jdminhbg 2y ago
I stand by my counterpoint: Google can’t run “Hey Google” on-device because it only controls a small % of the devices it’s available on. In the world of built-in voice assistants, Siri is behind, but only as behind as it was in 2021.
177.
by jdminhbg 2y ago
Depends on where you live: https://proofreadingpal.com/proofreading-pulse/writing-guide... .
178.
by jdminhbg 2y ago
This isn’t really the competition. This is more akin to opening up the OpenAI app than using Siri. Hey Google is, like Siri, still not LLM-based.
179.
by jdminhbg 2y ago
> The strategy will also save money. OpenAI spends about 12 cents for each word that ChatGPT generates because of cloud computing costs. There's a link on "12 cents" that goes to https://www.vantage.sh/blog
180.
by jdminhbg 2y ago
> isn't a tablet just a media consumption device? This is actually most of the storage space -- videos downloaded for consumption in places with no or bad internet.

More ›