epups

19 ms·

151.
by epups 3y ago
This article is misguided, because it compares self learning with the presence of a good teacher. For all kinds of reasons, there are places in the world where a child might be much better off watching a Khan Academy lecture than their loca
152.
by epups 3y ago
> To give a specific example, I find the idea that LLM based programming assistants will turbocharge software development to be based on hype not fact We already have empirical results suggesting this is not just hype: https:/&#x2F
153.
by epups 3y ago
That's a weird bar to clear. How many companies are offering a search engine that are profitable, "aside from 4-5 companies"? I don't see any fundamental technical problems at the moment, I see constant and tangible impr
154.
by epups 3y ago
What is the measured evidence that AI is hype right now?
155.
by epups 3y ago
I think skepticism is always welcome and we should continue to explore what LLM's can and cannot do. However, what I'm referring to is trying to get a quick win by defeating some inferior version of GPT or trying to apply a test w
156.
by epups 3y ago
If someone comes here and says "<insert programming language> cannot do X" and that is wrong, or perhaps outdated, don't you feel that the reaction would be similar? If you are trying to make categorical statements abou
157.
by epups 3y ago
Yes, hype exists and some things we thought were promising turned out not to be. However, if anyone is making the case that we know enough today to claim that AI is mostly hype, I think that's foolish.
158.
by epups 3y ago
I think ironically there has been an "AI-anti-hype hype", with people like Gary Marcus trying to blow up every single possible issue into a deal breaker. Most of the claims in this article are based on tests performed only on GPT-
159.
by epups 3y ago
The problem with your premise is that you're already drawing conclusions about the potential of AI and deciding it is hype. Perhaps decades ago someone could have equally criticised "Internet hype" and "mobile hype"
160.
by epups 3y ago
I think we should start talking about geoengineering again. It seems clear that the pace of what we are doing will not be sufficient to avert a disaster at this point. I think there are enormous risks involved with, say, stratospheric aeros
161.
by epups 3y ago
It's interesting, I feel excited as I haven't felt in a long time because wherever I look, there is a possible side project where I can explore AI and LLM's. On the other hand, I would feel very scared to try to transform any
162.
by epups 3y ago
Be it as it may, regardless of immigration policy we cannot watch and let people simply die in the sea. That is barbaric and it must be stopped.
163.
by epups 3y ago
If you would read the first link, you would see that what you're claiming is unsubstantiated. They could track it to a great level of detail because they identified the threat vector and patched it quickly.
164.
by epups 3y ago
I think you're resorting to quite a few strawmen here. Americans are not queuing up to the blood bank en masse, and I live in Europe and appreciate its standard of living. However I believe to have any chance of maintaining it, Europe
165.
by epups 3y ago
The problem is that to afford those social benefits, Europe inevitably also needs to care for its economy. Since tech is now the biggest economic sector and will likely continue to be for the foreseeable future, it is concerning if the tech
166.
by epups 3y ago
Microsoft knows which accounts were targeted by the attacker. They say so in the first link: "Our telemetry and investigations indicate that post-compromise activity was limited to email access and exfiltration for targeted users.&quot
167.
by epups 3y ago
The writeup by Microsoft is far more illustrative than the frankly confusing post and blog from the main article: https://www.microsoft.com/en-us/security/blog/2023/07/14/ana... Also, unlike wh
168.
by epups 3y ago
You can implement security measures on top of what is provided by Microsoft. If you have encryption at rest and you hold the keys locally, for example, even this high-level leak would not expose you. That said, good luck implementing and ma
169.
by epups 3y ago
"Security researchers agree" is a very broad statement. I don't believe there is a consensus at all. Fragmentation creates different problems than centralization, but it isn't a magical bullet either. Depending on your r
170.
by epups 3y ago
> Yes, but that doesn't preclude the fact that Twitter had 7500 employees prior to layoffs, without anything really to show for it. Well, they had 3x or 4x more income and market value than Twitter with fewer employees. That's
171.
by epups 3y ago
> We are rightfully condemning chattel slavery. However we can't do that if we are not appealing to a universal moral standard. We have no right to claim that the Southern states were wrong, if slavery in the way existed back then w
172.
by epups 3y ago
I don't believe it does, at all? How do you "solve" the moral problem of, say, whether we can breed and kill animals to eat them?
173.
by epups 3y ago
> How do human groups glorifying immorality imply that there is no universal moral code? If the universal moral code exists outside of those groups, then it can be applied to their actions, and we can call them wrong for the thing they g
174.
by epups 3y ago
If you had to qualify it with "most", then it isn't universal. Moreover, I don't think morality is what applies only to your friends.
175.
by epups 3y ago
There is a big difference between morality and science though. The physical reality around us does not depend on our opinion of it, and we can design experiments to test it. You cannot "prove wrong" a set of moral rules without ag
176.
by epups 3y ago
There almost certainly isn't a universal moral code. For every abominable act you can think of - murder, slavery, even genocide -, we can find examples of human groups that glorified them, usually by evoking religion, group affiliation
177.
by epups 3y ago
Maybe you could just have said "punks".
178.
by epups 3y ago
It is very possible, in fact likely, that gutting a ton of people directly involved in ad-related work - sales, press management, content monitoring, etc - had a negative impact on advertising revenue.
179.
by epups 3y ago
You are thinking too small here. Think about a system that could provide you with sophisticated image analysis to the level of a human expert. Tumor identification, quality control, etc. I don't think we are there quite yet, but this i
180.
by epups 3y ago
Yeah, I'm also skeptical about the actual value of specialists in the future. To me, AI brings a ton of power to generalists, who now have access to very powerful tools that would have taken them a long time to learn otherwise.

More ›