srveale

I think you misinterpreted my comment as "LLMs are good" which is a different conversation.

My point is: you can't say LLMs are a dangerous tool and call them vapourware at the same time. It's a contradiction in terms.

This places her on a spectrum where the farthest left you can go is the most left leaning US Democratic senator, which is not very "progressive" in the context of western politics as I mentioned in my last comment.

As someone who uses LLMs every day for general questions/brainstorming, more efficient coding, and building a product used by doctors to improve their documentation (saving them hours per week and freeing them to interact with their patients more personally), I would like more of this hot air please. Would you begrudge disabled people their new assistants? Language learners their translators? I could go on.

LLMs have some very important downsides, and I fully agree that they are dangerous, but we should dispel the notion that they don't have positive use cases. That leaves the benefits on the table, while the bad actors will continue with their destruction anyways.

Anyway, my original point was indeed just about the semantics of the word "vapourware", which if I'm interpreting the author correctly, would be better replaced with "malware" (not that I agree with such a stance).

In western politics, there are various definitions of the word "progressive". The definitions that include Kamala Harris are mostly used by right-wing Americans.

How much corporate funding did Bernie get?

Why do you think capital supported Kamala? Especially in hindsight?

And your joke about left vs right sponsorship of streamers has a very soft underbelly, which, if you don't know about it yet, kind of tells the whole story right there.

> The obsession with AI (and other vapourware) in our industry ... fuelling the hard-right — who coincidentally are very much using AI.

Is it useless or not? If it's vapourware, why would you care if the other side uses it? If the far right is using it successfully, then by definition it is not vapourware, right?

Shady business, potentially, but you might be underestimating how much some guys really, really need to have the most expensive watch in their friend group.

I was talking about people I know and talk with, mostly friends and family, who are smart, hard working, and their students are lucky to have them.

> Flipped classroom is just having the students give lectures, instead of the teacher.

Not quite. Flipped classroom means more instruction outside of class time and less homework.

> This is called "proctored exams" and it's been pretty common in universities for a few centuries. None of this addresses the real issue

Proctored exams is part of it. In-class assignments is another. Asynchronous instruction is another.

And yes, it addresses the issue. Students can use AI however they see fit, to learn or to accomplish tasks or whatever, but for actual assessment of ability they cannot use AI. And it leaves the door open for "open-book" exams where the use of AI is allowed, just like a calculator and textbook/cheat-sheet is allowed for some exams.

https://en.wikipedia.org/wiki/Flipped_classroom

IMO it's so easy to ChatGPT your homework that the whole education model needs to flip on its head. Some teachers already do something like this, it's called the "Flipped classroom" approach.

Basically, a student's marks depend mostly (only?) on what they can do in a setting where AI is verifiably unavailable. It means less class time for instruction, but students have a tutor in their pocket anyway.

I've also talked with a bunch of teachers and a couple admins about this. They agree it's a huge problem. By the same token, they are using AI to create their lesson plans and assignments! Not fully of course, they edit the output using their expertise. But it's funny to imagine AI completing an AI assignment with the humans just along for the ride.

The point is, if you actually want to know what a student is capable of, you need to watch them doing it. Assigning homework has lost all meaning.

I think the tricky bit is that AI companies make money off the collected works of artists, regardless of user behaviour. Suppose I pay for an image generator because I like making funny pictures in Ghibli style, then the AI company makes money because of Ghibli's work. Is that ethical? I can see how an artist would get upset about it.

On the other hand, suppose I also like playing guitar covers of songs. Does that mean artists should get upset at the guitar company? Does it matter if I do it at home or at a paid gig? If I record it, do I have to give credit to the original creator? What if I write a song with a similar style to an existing song? These are all questions that have (mostly) well defined laws and ethical norms, which usually lean towards what you said - the tool isn't responsible.

Maybe not a perfect analogy. It takes more skill to play guitar than to type "Funny meme Ghibli style pls". Me playing a cover doesn't reduce demand for actual bands. And guitar companies aren't trying to... take over the world?

At the end of the day, the cat is out of the bag, generative AI is here to stay, and I think I agree that we're better off regulating use rather than prohibition. But considering the broader societal impacts, I think AI is more complicated of a "tool" than other kinds of tools for making art.

True!

I get what you're saying, but I think "boiling frog" is more applicable than "chicken v egg."

You mention that people feel disappointed by ML systems because they don't feel intelligent. But I think that's just because they emerged one step at a time, and each marginal improvement doesn't blow your socks off. Personally, I'm amazed by a system that can answer PhD level questions across all disciplines, pass the Turing Test, walk me through DIY plumbing, etc etc, all at superhuman speed. Do we need neuroscience to progress before we call these things intelligent? People are polite to ChatGPT because it triggers social cues like a human. Some, for better or worse, get in full-blown relationships with an AI. Doesn't this mean that it "feels" right, at least for some?

We already know that among humans there are different kinds of intelligence. I'm reminded of the problem with standardized testing - kids can be monkeys or fish or iguanas and we evaluate tree climbing ability. We're making the same mistake by evaluating computer intelligence using human benchmarks. Put another way: it's extremely vain to say a system needs to be human-like in order to be called intelligent. Like if aliens visited us with incomprehensibly advanced technology we'd be forced to conclude they were intelligent, despite knowing absolutely nothing about how their intelligence works. To me that's proof by (hypothetical) example that we can call something intelligent based on capability, not at all conditional on internal mechanism.

Of course that's just my two cents. Without a strict definition of AGI there's no way to achieve it, and right now everyone is free to define it how they want. I can see the argument that to define AGI you have to first understand I (heh), but I think that's putting an unfair boundary around the conversation.

Not trying to be sassy but what definition of AGI are you using? I've never seen a concrete goal, just vague stuff like "better than humans at a wide range of tasks." Depending on which tasks you include and what percentage of humans you need to beat, we could be already there or maybe never will be. Several of these tests [1] have been passed, some appear reasonably tractable. Like if Boston Dynamics cared about the Coffee Test I bet they could do it this year.

[1] https://en.wikipedia.org/wiki/Artificial_general_intelligenc...

`alias | grep [alias name]` works as well, for me at least

Is there someone you're trying to disprove? LLMs are inherently statistical, as opposed to other techniques that rely on symbolic or logical relationships. I'm no expert but this is one of the very first things I learned when taking a class on neural networks.

"Here is the project description: [project description] Help me think of ways this could go wrong."

Copy some of the results. Be surprised that you didn't think of some of them.

"Rewrite the email succinctly and in a more casual tone. Mention [results from previous prompt + your own thoughts]. Reference these pricing pages: [URL 1], [URL 2]. The recipient does not appreciate obvious brown nosing."

I know nothing about this topic.

How crazy is it to cover the cables with passive sonar and detect damage threats? How much crazier is it to create a sufficient number of undersea drones that can prevent damage before it happens? Maybe manoeuvre a protective barrier over the predicted impact area if there's a dragging anchor or fishing net? Pick a fight with enemy drones?

I'm increasingly impressed and terrified with air/ground drone capabilities displayed in Ukraine. The sea floor seems like the next logical step. But maybe it's more efficient to detect damage quickly and make repairs easy.

This is what I meant in my comment. The seller needs to know what the buyer wants, maybe its a big bulk discount or an extra feature. If the seller decides the discount is too steep, or the extra feature doesn't fit with the road map, then no deal. Or maybe the deal really is that big, and it's worth catering to some one-off demands.

None of this means "hiding" information, but you can't put something like "We'll do X hrs of extra work if you buy Y licenses". Just like the any store might have a 10% discount if you buy a dozen, but if you want 50,000 then there will probably be a conversation involved.

What if you sell a product where it's easy to determine the cost for one user signing up by themselves, so you figure out the required markup and publish that on your site. But large organizations wanting licenses for each user will want a discount, will want finer details about contracts, and often some kind of unique adaptations to the product for their use case. The selling company needs to know if its worth the effort, in which case you have requirements gathering and negotiations. Of course there will be differential pricing depending on what the buyer company wants (cost goes up) and if it's a whale of a deal that the seller really wants (cost goes down) So... schedule a call?

Speciation would happen over hundreds or thousands of years. I doubt the notion of first world countries will outlast that time frame, even more so the 100% reproduction opportunity.

Isolated populations could come from space colonies, geographic isolation from war or extreme climate events, rich people with exclusive genetic upgrades, cyborg implants that make it possible to only reproduce with other cyborgs. Just off the top of my head.

If you're less patient, give it enough time the old fashioned way and homo sapiens as they are today would eventually not be able to reproduce with their descendants. Genetic drift.

Homo sapiens is subject to speciation just like every other animal.

Then there's cyborgs...

What do you think are the truest words ever spoken? Maybe just something out of the top ten.

Chiming in with helpful rules of thumb I go by:

1. Never give an estimate in the same conversation in which it was requested, unless you're quite sure the answer is "less than an hour". Especially if they are going to turn around and quote your timeframe to a customer. The reason for this rule is:

2. The more you think about an estimate, the higher it will go. "Oh yeah and we'll also have to..." is 100x more common than "I found a shortcut that doesn't compromise on..."

3. Even if they know it's an estimate and not a deadline, they might not know the difference between "40 hours of dedicated effort" and "One week's worth of business hours". Making that distinction is just as important as the estimate/deadline distinction.

I'm lucky to work where we don't spend much effort on estimates and time tracking, both of which just make things take longer and cause productivity-reducing stress.

Our invention works best except for extremely long flight times of 13 seconds

People's eyes glaze over when I rant about this. But yes, I fantasize about a world where you could say "let's set up a video call at 6" and there's no mental math when there's geographic distance involved.

Biggest downside I can think of, apart from the entire population groaning about making the switch, would be the confusion from reading literature from the before times. Oh, and all blood, sweat, and tears of all the devs who created timezone code going down the drain.

That sounds great, but I think it's more likely that they'll start charging for water and air before they start producing food and clothing with the intention of giving it away.

RTO usually means mandated, company wide RTO, not "there's an office if you want." So no, not right.

I can't see the comments, maybe because I don't have an account. So maybe this is answered but I just can't see it. Anyway: how can we be sure that this is the actual system prompt? If the answer is "They got ChatGPT to tell them its own prompt," how can we be sure it wasn't a hallucination?

"all of a given occupation should not be the same gender or race" is pretty obviously not equal to "a uniform distribution of races among every chosen subset"

> the betting markets

For context, there are 5 people and 1 bot betting, and it's not real cash or even something with a cash equivalent, and the bet amounts are less than what you get for free on sign up -- not exactly what I'd call a "market". And the bet is that they'll close Google Cloud in 2 years, not 5-10 years as is being discussed.