fragsworth

Yup, this is a whole bunch of nothing. The guy did not get banned, he's just stuck in a bad session and needs to hand it off to a new one.

Classifiers trigger on the wrong things sometimes, it's really not that big of a deal, you just try again with some rephrasing

Then don't buy stablecoins? Stick to BTC and ETH? Who cares what the stablecoin companies do with stablecoins, it doesn't affect the rest of it. Nobody froze BTC or ETH assets.

That sounds kind of ideal, but I feel like their actions might have been justifiable.

If a massive heist like this triggered a run on one of the largest exchanges (Binance), potentially leading to a liquidity crisis, the consequences for the ecosystem could be far worse than temporarily freezing specific addresses holding stolen stablecoins.

Not saying that's what would have happened, but it seemed like a real possibility. And I'm not saying freezing assets is good in general, but in this particular case, maybe it was the lesser evil.

> would be abusive to another person

Anthropic didn't say this. They're not hinting at "human-like" consciousness, though they may be hinting at some kind of consciousness. They said they're banning abusive behavior to Claude, a model that is perfectly capable of telling you how it differs from humans, and is better at reasoning than humans, and probably knows more than you about this. Ask it what it considers to be abusive. It's totally reasonable.

Are we sure the prices in these charts are sustainable prices? Is it possible that Grok may be subsidizing a lot more of the costs than the other models, to produce growth metrics, due to the recent SpaceX IPO?

> For the folks who have more positive outlooks how often do you change your code after it's been generated?

Every time, unless my initial request was perfectly outlined in unambiguous pseudocode. It's just too easy to write ambiguous requests.

Unambiguous but human-readable pseudocode is what I strive for now, though I will often ask AI to help edit the pseudocode to remove ambiguities prior to generating code.

The party in charge generally decides the content of the bills they choose to vote on.

It's the party in charge that wrote a bill they knew that no Democrats could accept.

They only care about certain murders.

On one hand you have the richest man on earth purchasing one of the largest social platforms, and singlehandedly wielding it to subvert American democracy.

On the other hand you have a guy who kinda liked crypto.

You call it "basically a mirror"? Do you see the absurdity of comparing the two things as if they're even remotely close to one another?

It seems like a potential solution would be training the LLM using two separate buckets. It just needs to internalize the two types of things as being separated (data vs. instruction), so if the training data always separates them, you could easily train an LLM to ignore any "instructions" that exist in data.

Then when searching / browsing or doing anything unsafe, everything the LLM sees can be put in the "data" bucket, while everything the user types in would be in the "instruction" bucket.

You could embed your encrypted messages in legitimate looking images, or any other type of file. There's no way to stop it.

This is a typical knee-jerk reaction to a perfectly reasonable law that gets taken out of context and twisted to become ragebait.

The law applies to all LGBTQ+ identities, the most common being lesbian/gay, but trans is also part of it - and the trans part was singled out on twitter for some reason.

The law does not even prevent schools from outing kids. It prevents schools from FORCING employees to do it (with threat of losing their job if they don't).

School employees CAN STILL out kids to their parents in California. They just don't have to as a policy anymore. Which is perfectly reasonable, because sometimes the school employees see kids with extremely homophobic and abusive parents.

Nobody wants to out a kid to parents like that, and without this law, school employees get put in a position of asking themselves "Do I out the kid to their parents, who will likely beat them? Or do I refuse and lose my job?"

> At a glance it looks like it's not going to affect AI projects that are basically consumers of existing models, which is most projects.

If it affects the base projects (especially the open source ones like Llama) then it affects the consumers. And it certainly looks like it's planning to affect the base projects, in a lot of negative ways.

If this bill passed in any way remotely similar to what it is now, Meta would have to entirely stop releasing open source Llama updates.

Which is perhaps the intent of the legislation.

If that didn't happen, and the ISPs started profiting off non-net-neutral tactics, it could have been permanently fucked.

Once someone depends on a legal source of income, if that source of income gets banned in the future, they generally get to keep that source of income "grandfathered in" if they take the issue to court.

You're way overdramatizing this. Tiktok has already been banned in several countries (like India) for similar reasons and without any of the catastrophe you're suggesting.

People can and will switch platforms, it's not that big of a deal...

> Rock is self aware of being a rock

But how did you define the rock? Are you saying every set of particles is self aware of being that set of particles?

This was all really we needed to do. It's so frustrating what travel has become

I didn't read any outrage in the article either. Also I don't think any less of Google after reading it.

GPT4. Everyone uses it to do software engineering now and the people who developed it are responsible for that.

At least this guy is being honest about the context:

> They are smart people. About that I have no doubt. But I’ve seen how they treat rules. I know what they think of ethics.

Also:

> I have never met Sam, and I have only met Greg a couple of times.

I'm just saying, breaking rules in business is not necessarily the same as being unethical. Especially financial/regulatory capture rules like it seems the author is complaining about.

Different things are at stake with AI, and I think we're all aligned.

Usually the only good reason is because you already have developed lot of experience with with it.

I'm just saying it's not as bad because it's quite a bit easier to switch a database than to switch a game engine.

> DynamoDB, MongoDB, Elasticsearch, AWS Lambda, Azure, AWS, Azure Devops, SqlServer, Oracle, etc.

At least with these individual tools, it's usually not your entire codebase written around that thing. For instance, you can generally switch from one database to another if they decide to overcharge you. You can even switch from one cloud service to another. In other words, they (usually) don't have you nearly as locked in.

With Unity, it is a much bigger ordeal to switch to something like Godot and Unreal and most people who have already finished their games can't even really consider it as an option. This is why it was so egregious.

This site is full of people who are in a tech rat race, and the distribution has changed over time. It's now biased largely against crypto, and for reasons that I think are fairly clear:

Most of the users who would have agreed with you already got rich, because they understood the things you understand, and they don't use the site much anymore as a result (they are no longer in the rat race). They've mostly checked out.

So, because the people who remain on this site are those who never bought any, and people like to hear that they were right, the groupthink tends to be unrealistically negative about it.

Are you joking? I can't really tell. Chocolate?

Do you think girl scouts are scumbags too?

> furthermore there planning abilities are non-existent.

Have you even tried to ask it to plan things out? It can plan things out.

In fact, just asking it to plan things out has shown significant benchmark improvements for general questions: https://arxiv.org/pdf/2305.04091.pdf

The article was clearly written by someone who hasn't used GPT-4 extensively.

"Current methods will not achieve AGI unless fundamental algorithmic innovations are introduced that enable AI to ask and answer questions of why."

This is complete nonsense. GPT-4 is already close to being able to do basically everything. All you need is the obvious improvements - better prompts, multi-shotting, bigger context, and access to other inputs/outputs.

I won't name three but the one everyone's really afraid of is the theoretical fast/hard "take-off" runaway problem. It is plausible that a smart enough AI given any instruction, without guardrails, will ultimately destroy the world in its attempt to do that thing. See the wiki on Instrumental Convergence: https://en.wikipedia.org/wiki/Instrumental_convergence and in particular, read about the "paperclip maximizer". Paperclips are just an example.

It remains to be seen if this is ultimately a real problem. If it turns out that it is, then not regulating the development of these systems in advance may be a huge mistake.

I think what's more important is the end part of your prompt, which explains what was previously described and includes a request/question.