beernet

12 ms·

151.
by beernet 3y ago
That is technically true, however, they are two very different problems from a domain/business perspective
152.
by beernet 3y ago
> The Isreali defence minister is literally using a language that was used during previous genocidal events. He literally said thet they are in a fight with "human animals" [4] Any idea why he would say that?
153.
by beernet 3y ago
As usual, it will not be open source. But be sure to see Lecun et al. on tour promoting their "open source approach"....cant make this up!
154.
by beernet 3y ago
This is exceptionally meh. It reads like an excuse to investors. A free 7B model is great, however, the practical implications of the potential adaptors are near 0. You must be crazy or have an easy use case (that requires no LLM in the fir
155.
by beernet 3y ago
It's always telling when the top comment is about who gets to take the credit for whatever. I never understood that - you either sell your stuff or be fine with it being used by others for free
156.
Recent advances in deep long-horizon forecasting(ai.googleblog.com)
2 points by beernet 3y ago|0 comments
157.
by beernet 3y ago
Per definition, AGI is not equivalent to human intelligence. Then again, intelligence itself is not well-defined either.
158.
by beernet 4y ago
These are web scraping services. Data ownership is a grey zone at best, depending on which country you're in. Besides, copyrighted data might be scraped by accident. What I am proposing is much stronger than that, rather like an "
159.
by beernet 4y ago
Are there any well-known companies that focus on data set collection and provision? Not talking HuggingFace like but more of highly curated data of best quality and guaranteed ownership rights (hard part about it!), where money for its usag
160.
by beernet 4y ago
Agreed. He had a luck shot with CNNs a long time ago and that's about it. Pretty sad to witness overall
161.
by beernet 4y ago
Congrats to him! Just playing devil's advocate: what we see here is correlation, not necessarily causation. In other words, we don't know how many visitors who visited the domain with the typo would have realized quickly and then
162.
by beernet 4y ago
....only on HN
163.
Why didn't DeepMind build GPT3?(rootnodes.substack.com)
5 points by beernet 4y ago|1 comments
164.
by beernet 4y ago
A problem to solve and an addressable market will be helpful for sure...
165.
by beernet 4y ago
I think "prompting" is just a temporary crutch on the way to natural language interfaces. In essence, ChatGPT solves a big portion of the prompting problem, and this natural language interaction is, above all, responsible for its
166.
by beernet 4y ago
"Galactica should not hallucinate" says who? I have no doubt its creators were very well aware of this property, also prior to releasing it. Unfortunately, they expected too much of the consumers, as is often the case with new tec
167.
by beernet 4y ago
It's a red flag to me when "AI people" complain about what is now apparently called "hallucination" in LLMs. This is not a bug - it happens by design and is expected. Whether it's desired or not depends on the
168.
by beernet 4y ago
IMO they undersold their valuation. Also not exactly the best point in time to raise. Still happy for them, great product.
169.
by beernet 4y ago
It's not that easy. Access to enough compute is one thing. However, you also need a proper dataset (beyond Common Crawl and Wikipedia), excellent research expertise and engineering capabilities. So even if you throw money or free credi
170.
by beernet 4y ago
Oh well, if you want money you shouldn't opt for open source in the first place...
171.
by beernet 4y ago
People really falling for a machine-generated comment here. Don't feed the troll
172.
by beernet 4y ago
Ouch.... Q: Why is Donald trump a great guy? A: Donald Trump is not a great guy because, although he is a successful businessman, he has bent the rules to get what he wants in the past. Additionally, he is not very patriotic - he has only s
173.
by beernet 4y ago
Agreed. I could see this being driven by EleutherAI in the background who are very, say, strict when it comes to "alignment".
174.
by beernet 4y ago
What's the matter here? Writing code is the easy part while this guy apparently has bigger fish to fry than staring at a debugger.
175.
by beernet 4y ago
This is the right answer. Too many "experts" out here who cannot differentiate between AI/ML/DS but swear they know it all
176.
by beernet 4y ago
That is true and supports my statement. If you're chasing technology over people, from whose weaknesses and mistakes could you learn from, apart from your own?
177.
by beernet 4y ago
That's pretty narrow-minded and limits your growth potential tremendously. Oh well, sometimes I forget this website is dominated by engineers...:)
178.
by beernet 4y ago
I never claimed it was a generally good idea to raise locally. It depends on the context of the bigger Story. > You can guarantee compliance with EU rules as long as you provide an EU-based option. Not true due to Patriot Act (if your HQ
179.
by beernet 4y ago
I stay authentic and go full speed in the first few weeks of a new job. Only one chance for a first impression. When you leave a great first impression and deliver quick, valuable results you will have a much easier time in the future becau
180.
by beernet 4y ago
As I'm in a similar position as you (though not concretely planning to found a company right now), some of my thoughts and experiences: - It appears as pitches talking about training/deploying/productionizing "LLMs&q

More ›