Same reason we study and do research in biology and fields in general. Understanding more about biology, whether in neuroscience or elsewhere, is the fundamental backbone of medicine. Do you want to live in a world where we don't pursue scientific research into things because they seem scary to you? Does it not concern you us building AI smarter than humans, while knowing little to nothing about our own brains? Before we start fading in relevance, we ought to understand how our own bodies (and minds) work! So in fact I think the exact opposite. We should be doing much more work in understanding our own biology (as long as it's ethical).
At 8x86B, looks like the largest open model yet by far. Would be interesting to hear how many tokens it's been trained on. Especially important for higher param models in order to efficiently utilize all those parameters.
He's not asking for money in the lawsuit, so what is it going to do? I thought it was for show in the first place, the goal was to hold them to their founding principles. Which are of course, pretty subjective. Top to bottom it's God complex over there and the secret justification for closed source was pretty ironic--we only found out this key unmentioned detail once the lawsuits start flying.
It was his money to give, he's not obligated to giving it out without some form of control. But yes that does seem pretty excessive, he basically wanted to pull a Tesla there.
Anyone can sue anyone for any reason at any time. Unless the lawsuit is in bad faith and meritless (only a court can make that determination), there is no argument to be had about "lawfare". The only difference here and in other companies is that Nintendo actually follows through with their legal threats instead of just filing complaints they don't intend to take to court.
LLaMA 3 with >=70B params will be launching this year, so I don't think this is something that will hold for long. And Mixtral 8x7B is a 56GB model, sparsely. For now I agree, for many companies it doesn't make sense to open source something you intend to sell for commercial use, so the biggest models will likely be withheld. However, the important more thing is that there is some open source model, whether it be from Meta or someone else, that can rival the best open source models. And it's not like the param count can literally go to infinity, there's going to be an upper bound that today's hardware can achieve.
The 5th Circuit court of appeals has already ruled similarly that the SEC's in house courts are unconstitutional, and that Congress can't delegate its legislative/judicial powers to the executive branch in that case under the separation of powers [1], a case that will likely be heard by the SCOTUS. By extension, if that is upheld, the idea is that this same principle could be applied to other executive branch administrative courts like in use at the NLRB.
I love it when people make this “nuke” argument because it tells you a lot more about them than it does about anything else. There are so many low information people out there, it’s a bit sad the state of education even in developed countries. There’s people trotting around the word “chemical” at things that are scary without understanding what exactly the word means, how it differs from the word mixture or anything like that. I don’t expect most people to understand the difference between a proton and a quark but at least a general understanding of physics and chemistry would save a lot of people from falling into the “world is magic and information is hidden away inside geniuses” mentality.
Why not just use an LLM to generate the code/commands for you? Making a whole new language someone has to learn and memorize special keywords and built-ins doesn't seem too appealing by today's standards.
It’s not really a vulnerability in the sense that it leads to any sort of system compromise. It’s definitely a design flaw in whatever features they added to the OS, but not necessarily something that warrants a huge investigation.
This is why I am concerned that really good AI tools (be it LLMs, or not), are only going to make people dumber over time. It is true that those people that are highly motivated will be digitally enhanced by LLMs, but the hard reality is that this does not encompass most people. Most people will do the least amount of work possible to solve a task. Rather than taking the time to understand and critically reason about the output of an LLM, these people will instead take it and run, only asking questions later [if at all].
And as we zoom out and think more farther into the future, I see it getting much worse. If AI is really doing all the "hard stuff", then the general human incentive to learn and do them at all quickly treads to 0. This isn't going to be everyone, I think some people will absolutely become "10x developers" or the equivalent for other domains. But all this will result in is more inequity, in my naïve view. The universe has a fundamental property that things move from higher energy states into lower ones. From a human POV, I think you could apply a similar idea, if the need to be smart quickly recedes, then in general we may degrade unaugmented human collective intelligence over time.
I don't know, maybe we'll figure something out to make it much easier to learn and ingest new concepts, but it seems more and more to me that the high obstacles for human brains learning things (with poor memory) is too big a bottleneck to overcome any time soon.
People should never have blindly trusted anything on the internet in the first place. As far as I’m aware, most of the fears like this are overrated. People will live, learn and adapt as we always have. Constantly complaining about doom and having pessimism about anything seemingly scary is not a great way to live a fulfilling life.
The Hacker in hacker news was never meant to imply black hat/malicious types of hacking. There’s quite a difference between say, tearing something apart, reverse engineering it, breaking into something that _you_ own, versus trying to tear into something you don’t own without a really good reason. At the end of the day it’s about judgement and taste, there isn’t so much a hardline but there is a line on what we consider acceptable and unacceptable areas of exploitation. Beyond the piracy point, I think few could find this exploitative, it seems like a cool open source project that could genuinely offer better and customizable user experience.
This sounds even worse for Apple. If it’s too easy to make an in-app purchase, why don’t they make it harder to complete one by adding password requirement for payments through their App Store? Of course they won’t do that because it hurts /their/ bottom line. Following this logic, letting people complete payments outside of Apples control should absolve Apple of the legal liabilities between payments made between the user and a third party. If this is not the case, can you sue Apple for unauthorized payments made over Safari?
By "doing something for 10 years", I think the implication is you're actively working on that subject area, regularly, for 10 years. Not per se occasionally working on something every now and then. You can become a really good Chess player if you consistently play for 10 years--you figure out all the strategies and shortcuts, and it would be unusual if you weren't really good compared to someone playing for a year (assuming they're not some exceptional learner). But if your metric is just "I've played it occasionally since 10 years ago" then indeed, you probably didn't develop much breadth or scope as you were constantly forgetting and relearning as opposed to compounding your knowledge.
As for programming, I don't think programming is all that hopelessly complex and broad as field, but it can seem that way to a beginner. Most computer science concepts translate very well to other parts of the field, and the core programming constructs and libraries don't change much at all. How many ways can you configure a website, a mobile app or a database? Your instinct might be to think about all the different libs you can pull in, all the different programming languages, etc. But they all do roughly the same thing, they all compile down to the same stuff. You just have to develop the skill of understanding the fundamentals as opposed to getting lost in a sea of high level abstractions.
A lot of attention gets paid to sex abuse content, which it should, but there is much less I feel for all sorts of other abhorrent content. Gore, violence (physical/mental), terrorism, death, destruction, shock content, etc. It's not just NSFW, but also NSFL content that I think is under discussed at times. Seeing the worst output of humanity non-stop can (and _will_) completely break you mentally if consumed at a sufficient amount, as we are at the end of the day humans. It's unfortunate the world is a messed up place, and maybe it will stay that way as long as humans are around.
This is the exact type of content that I think AI is so crucial for detecting, and it's a bit sad that we often hear so about all the bad things you can generate with AI and so little about all the good. If we need to train models on this type of content to make it easier to detect and remove, to prevent mentally scarring people, then personally I'm all for it -- regardless of the "it can also generate that bad content" cost.
I assume this works like traditional cellular roaming, no? So this would mean you have to be a customer of one of the service providers to get reciprocated roaming capabilities abroad.
I don't think it's that unusual. What comes to mind immediately is it's not unusual for me to clone something from a private git repo, where a username+password would be needed for permissions. In which case it's possible to put in `git clone http://username:password@example.com` or another git command that interacts with remotes. (To be clear the "password" is typically a token and not human generated string, but still functions like a password).
This particular rejection vote by the US in the UN is a bit peculiar to me. Obviously, I understand the sensitivity around the situation--so why not abstain from the vote? Why when given that option, not use abstention and instead choose rejection? It sends a powerful and deliberate message, and I'm not sure who it's supposed to appease nor what interests that represents.
Why would normal memory access be O(log n)? Seems to me jumping to a random address should be an O(1) operation. Would be pretty dumb if jumping to higher RAM addresses takes longer than jumping to lower RAM, excluding hardware things like different RAM sticks or page files.
This seems pretty incoherent to me. It’s a hodgepodge of complaints about all sorts of things like bad code, slow job running, version control and other stuff. You can say the same things about any complex software. Not sure what the takeaway is supposed to be, as they are not offering any suggestions or alternatives or other products to compare GitHub Actions with. For a free tool, I think the opposite, it’s one of the best CI services I’ve ever used, and you really don’t get perspective until you’ve used something worse.
It’s the “least controversial idea” if you have faith in the government to do the right thing. Many people don’t, for one reason or another. Fortunately, more people (and companies) are realizing that liabilities like this can be absolved by simply not having the data, or access to the data, in the first place. This involves storing data locally on client when possible, using end to end encryption when possible, encryption by default (the default on latest HTTP, DNS, etc), and simply not holding sensitive data in the first place. There’s still work to do, but we are moving in the right direction.
No, I disagree with the idea that one letter domains look wrong. Especially with the many new TLDs, more often than not short URL services use single character domains. Take g.co for example—google uses that in ads everywhere. It’s modern equivalent of calling a short code phone number unrecognizable. But yet most people still instinctively know how to text them for information. However for really esoteric TLDs, I can see why it would be more of a problem, not because the length of the domain, but that the TLD, like .jobs, just doesn’t sound right.
Tit for tat moves like this serves no legitimate public interest. The fact that Nvidia can even "redesign" and nerf chips on the fly should perhaps make it obvious to keen observers how ridiculous trying to control compute is. The bureaucrats have no idea what they're working with, a surface level grasp of technology and then naturally confabulate how impactful these trade wars are for political clout. If it were China doing this against the United States, what do you think would happen? Would we just bend over and deal with the slow tech? Or would we be pushed to work even faster? Obviously, all this is going to do is spur up domestic ASIC chip design and production. It's a net-win for China, with some short term pain.
Completely agree. When I talk to others about Github Copilot/ChatGPT however, I am rather surprised to hear some of the criticisms. As in, it's unreliable, it's wrong, not good for beginners, etc. I don't understand that criticism at all. Yes, it's occasionally, perhaps frequently wrong, but so is virtually any other source of information. What's the alternative? Well, you Google your way around half a dozen websites, read through forum posts, bloated/incomplete doc websites, Q&A sites, etc. It's absolute hell to me, stuff that can take 10-15min on average to deal with, so comparatively using LLMs to not is like light and day. Maybe it's due in part to people being inexperienced and lacking the ability to independently judge whether content seems right or not, so I will admit you do need to be a bit of an expert to fully utilize these tools.
But even as a beginner--when I started coding some 10 years ago, I'd have killed for something like ChatGPT. Had some programming problem you needed to solve? You better hope that someone wrote something about it online, that it's on StackOverflow or been discussed on some other discussion board. Otherwise, it's up to you to splunk into StackOverflow, wait a week or two, hope you don't get ignored/downvoted, or try to find some IRC channel to post the question to. Comparatively having the ability nowadays to talk to a AI about the most niche programming concepts in your specific use case, in English, without being judged? It's straight up magical to me.
I don't think you understand where those comments are coming from. I didn't interpret it as malicious but rather venting his frustration. It's obviously (very) inappropriate, but there is a (very) big difference between speaking out of frustration and hate. And the apology was after, not before that remark, to be clear. If you have it out for yourself as a Musk-hatee, for one reason or another, I don't think it's meaningful to argue on the semantics. It's not going to win back advertisers, but that's a problem for X/Twitter to deal with, not me. But IMO, and just my view, he's done what is reasonable in apologizing for his remarks, and sometimes you can't do much more but just move on. Likewise, there is nothing of value to extract from these types of debates, hence my last point on focusing on meaningful actions over drama.
It was an emotional response driven from frustration. But obviously something highly inappropriate to say outright, and only creates yet another PR disaster. He is his own worst enemy, unable to adequately control himself. It seems verbatim what Dana White said the other day. That said, he did apologize for the remarks and the rest of the interview was rather normal. We'd all be better off not clinging to the constant desire for drama and focusing on meaningful actions over words.
As an FYI, Bing Chat at http://bing.com/chat continues to work even during OpenAI outages. It's also running GPT-4 -- it can be annoying when it reaches out to search, but you can usually explicitly prompt it to not do that.