With NHS mental health waitlists at record highs, are chatbots a possible solution?

taking Betteridge’s Law one step further - not only is the answer “no”, the fucking article itself explains why the answer is no:

People around the world have shared their private thoughts and experiences with AI chatbots, even though they are widely acknowledged as inferior to seeking professional advice.

as with so many other things, “maybe AI can fix it?” is being used as a catch-all for every systemic problem in society:

In April 2024 alone, nearly 426,000 mental health referrals were made in England - a rise of 40% in five years. An estimated one million people are also waiting to access mental health services, and private therapy can be prohibitively expensive.

fucking fund the National Health Service properly, in order to take care of the people who need it.

but instead, they want to continue cutting its budget, and use “oh there’s an AI chatbot that you can use that is totally just as good as talking to a human, trust us” as a way of sweeping the real-world harm caused by those budget cuts under the rug.

Nicholas has autism, anxiety, OCD, and says he has always experienced depression. He found face-to-face support dried up once he reached adulthood: “When you turn 18, it’s as if support pretty much stops, so I haven’t seen an actual human therapist in years.”

He tried to take his own life last autumn, and since then he says he has been on a NHS waitlist.

tl;dw is that you should say “please” as basically prompt engineering, I guess?

the theory seems to be that the chatbot will try to match your tone, so if you ask it questions in a tone like it’s an all-knowing benevolent information god, it’ll respond in kind, and if you treat it politely its responses will tend more towards politeness?

I don’t see how this solves any of the fundamental problems with asking a fancy random number generator for authoritative information, but sure, if you want to be polite to the GPUs, have at it.

like, several lawyers have been sanctioned for submitting LLM-generated legal briefs with hallucinated case citations. if you tack on “pretty please, don’t make up any fake case citations or I could get disbarred” to a prompt…is that going to solve the problem?

short answer: no, not really

long answer, here’s an analogy that might help:

you go to https://yourbank.com/ and log in with your username and password. you click the button to go to Online Bill Pay, and tell it to send ACME Plumbing $150 because they just fixed a leak under your sink.

when you press “Send”, your browser does something like send a POST request to https://yourbank.com/send-bill-payment with a JSON blob like {"account_id": 1234567890, "recipient": "ACME Plumbing", "amount": 150.0} (this is heavily oversimplified, no actual online bank would work like this, but it’s close enough for the analogy)

and all that happens over TLS. which means it’s “secure”. but security is not an absolute, things can only be secure with a particular threat model in mind. in the case of TLS, it means that if you were doing this at a coffee shop with an open wifi connection, no one else on the coffeeshop’s wifi would be able to eavesdrop and learn your password.

(if your threat model is instead “someone at the coffeeshop looking over your shoulder while you type in your password”, no amount of TLS will save you from that)

but with the type of vulnerability Jellyfin has, someone else can simply send their own POST request to https://yourbank.com/send-bill-payment with {"account_id": 1234567890, "recipient": "Bob's Shady Plumbing", "amount": 10000.0}. and your bank account will process that as you sending $10k to Bob’s Shady Plumbing.

that request is also over TLS, but that doesn’t matter, because that’s security for a different level of the stack. the vulnerability is that you are logged in as account 1234567890, so you should be allowed to send those bill payment requests. random people who aren’t logged in as you should not be able to send bill payments on behalf of account 1234567890.

oh, this one’s pretty easy, actually

a normal AI tells you it’s safe to eat one rock per day

an AI agent waits for you to open your mouth, and then throws a rock at your face. but it’s smart enough to only do that once a day.

Casey Newton reviewed OpenAI’s “agent” back in January

he called it “promising but frustrating”…but this is the type of shit he considers “promising”:

My most frustrating experience with Operator was my first one: trying to order groceries. “Help me buy groceries on Instacart,” I said, expecting it to ask me some basic questions. Where do I live? What store do I usually buy groceries from? What kinds of groceries do I want?

It didn’t ask me any of that. Instead, Operator opened Instacart in the browser tab and begin searching for milk in grocery stores located in Des Moines, Iowa.

At that point, I told Operator to buy groceries from my local grocery store in San Francisco. Operator then tried to enter my local grocery store’s address as my delivery address.

After a surreal exchange in which I tried to explain how to use a computer to a computer, Operator asked for help. “It seems the location is still set to Des Moines, and I wasn’t able to access the store,” it told me. “Do you have any specific suggestions or preferences for setting the location to San Francisco to find the store?”

they’re gonna revolutionize the world, it’s gonna evolve into AGI Real Soon Now…but also if you live in San Francisco and tell it to buy you groceries it’ll order them from Iowa.

click here to pre-order my upcoming book, published by Harvard Business Review, “Don’t Be A Fucking Nazi and Other Secrets To Corporate Success”

their pricing page is here.

I’m paying 10 USD/month for their unlimited plan, there’s also a 5 USD/month tier but I’m sure that I would exceed its 300 searches/month limit.

so it’s not dirt-cheap, but not stupidly expensive either. I can afford it, and I’m happy to pay it because it’s a business model that I would like to see succeed.

I’ve been using Kagi for ~2 months, after a friend gave me a similar invite code. this news from Google affirmed my decision to pay for Kagi once the 3-month trial is over, instead of going back to Google.

Tay any% speedrun

although I suppose “only one day after launch” doesn’t break the record:

It caused subsequent controversy when the bot began to post inflammatory and offensive tweets through its Twitter account, causing Microsoft to shut down the service only 16 hours after its launch.

but it’s great that the billionaire owner of the LA Times is trying. this sort of innovation is why billionaires like him are so important.

putting nukes into space is quite unlikely, even taking into account the current clusterfuck of the US government.

it’s been thoroughly studied since the 1950s, for obvious reasons. the practical considerations put it somewhere between “not feasible” and “gigantic pain in the ass”.

nuclear weapons need maintenance and upkeep, which the US military is already not terribly good at. a large part of this is that during the Cold War, maintaining nukes was seen as an important job within the military. in the past few decades, if you want career advancement in the military, you’d want to go to Iraq or Afghanistan for actual combat. working with nukes has become somewhat of a dead-end, career-wise.

satellites in LEO have a finite lifespan - the tiny bits of atmospheric drag mean they need to spend a bit of fuel to maintain altitude. after the fuel runs out they’re de-orbited, usually into the south Pacific (one of the most believable theories about the purpose of the X-37 space plane is refueling CIA spy satellites). doing that with nukes would be extremely expensive, as well as environmentally catastrophic (though of course the current government would only really care about the former)

and on top of all that…the US simply doesn’t need nukes in space. there is the “nuclear triad” of land-based ICBMs, nuclear-armed bombers, and nuclear-armed submarines. that was established during the Cold War to ensure the US had the ability to strike back at Russia, even if Russia devastated the US with a first strike.

the more realistic scenario in my mind is Kessler syndrome - a satellite-on-satellite collision creates debris, and that debris takes quite a while to fall out of orbit. in the meantime, it can create a chain reaction by colliding with other satellites. space is big, but LEO is much more crowded than it used to be, particularly with Starlink satellites, and those are cheaply manufactured and don’t always have reliable thrusters to allow them to move out of the way of any debris.

In the first half of 2024, satellites belonging to SpaceX’s Starlink fleet performed almost 50,000 collision-avoidance manoeuvres.

if it did happen, Kessler syndrome wouldn’t have much of an immediate impact, but instead a longer, slower-burning one. launches of new satellites into LEO would become less frequent due to the increased risk, and higher orbits (GPS and geosynchronous satellites) would be more risky as well because they would need to pass through the debris cloud. so existing satellites would continue to work, but as they aged out and needed replacement, those replacements would be less likely to happen.

https://en.wikipedia.org/wiki/Nazi_archaeology#The_Search_for_Atlantis

In their search to prove the superiority of the Aryan race, the Nazi party began searching the world for archeological evidence that would prove to the rest of the “inferior” world that the German people were not only a superior race, but that they transcended traditional human standards. One archeological exploit made popular by the movie Raiders of the Lost Ark was the search for the Ark of the Covenant. Another, perhaps less known, exploit was their attempt to discover the lost island of Atlantis.

yeah, it’s a real mystery why Rogan is surrounded by people who believe in Atlantis.

a salary that guarantees $1 million a year post-tax

to keep the mind-boggling numbers in perspective:

you’re paid $1 million/year post-tax, like you said.

and say you have no expenses to speak of - you take all your meals in the Google cafeteria, take the Google shuttle to work, and live with your parents or in some other form of housing that doesn’t cost you anything. this means you can put that entire $1 million/year into a savings account.

even in that contrived scenario, you would need to work 1000 years to accumulate one billion dollars.

at which point, you would have 1/145th of Sergey Brin’s current wealth. if you wanted to match it, you would need to work 145,000 years.

here is the original source of the article, published on a site called Futurism: https://futurism.com/microsoft-ceo-ai-generating-no-value

it got syndicated by Yahoo News because Yahoo does a ton of that in a increasingly desperate attempt to be relevant

judging by the “more top stories” on Futurism’s home page right now, they lean pretty heavily on clickbait:

Trump White House Tells Elon He’s Stepped Over the Line

Microsoft Backing Out of Expensive New Data Centers After Its CEO Expressed Doubt About AI Value

Shark Steals Camera, Capturing Amazing Footage From Inside Its Mighty Jaws

here is the primary source that the article is based on: https://www.dwarkeshpatel.com/p/satya-nadella

there’s a transcript that I suspect is almost certainly AI-generated, so some of these quotes may not be completely accurate:

Satya, thank you so much for coming on the podcast. So just in a second, we’re going to get to the two breakthroughs that Microsoft has just made. And congratulations, same day in nature. Majorana Zero chip, which we have in front of us right here, and also the world human action models.

right off the bat, we have the context that this is a friendly interview for Nadella to promote some new “breakthroughs” that Microsoft has. this may be explicit spon-con or just “regular” access journalism, it’s hard to say.

around 15 minutes in, the host asks:

You recently reported that your yearly revenue from AI is $13 billion. But if you look at your year-on-year growth on that, in like four years, it’ll be 10x that. You’ll have $130 billion in revenue from AI if the trend continues. If it does, what do you anticipate… we’re doing with all that intelligence?

Like this industrial scale use, is it going to be like through office? Is it going to be you deploying it for others to host? Is it going to be, you got to have the AGIs to have 130 billion in revenue? What does it look like?

and Nadella responds:

Yeah. I see the way I come at it, Dworkish, is it’s a great question because at some level, if you’re going to have this sort of explosion, abundance, whatever commodity of intelligence available, the first thing we have to observe is GDP growth, right? Before I get to what Microsoft’s sort of revenue will look like, I mean, there’s only one governor in all of this, right? Which is, this is where a little bit of, we get ahead of ourselves with all this AGI hype, which is, hey, you know what? Let’s first see if, let’s say develop, I mean, like, remember, like, the developed world is what? 2% growth, and if you adjust for inflation, it’s zero? That, like, so in 2025, as we sit here, I’m not an economist. At least I look at it and say, man, we have a real growth challenge. So the first thing that we all have to do is let, and when we say, oh, this is like the industrial revolution, blah, blah, blah. Oh, let’s have that industrial revolution type of growth. That means to me, 10%. 7%, developed world, inflation adjusted, growing at 5%. That’s the real marker, right? So it’s not just, it can’t just be supply side, right? It has to be, in fact, that’s the thing, right?

I think there’s a lot of people are writing about it. I’m glad they are, which is the big winners here are not going to be tech companies. The winners are going to be the broader industry that uses this commodity that, by the way, is abundant. Suddenly, productivity goes up and the economy is growing at a faster rate.

When that happens, We’ll be fine as an industry. But that’s, to me, the moment, right? So it costs self-claiming some AGI milestone. That’s just nonsensical benchmark hacking to me. The real benchmark is, is the world growing at 10%.

that word salad is a lot of things, but I don’t think it lives up to the “generating basically no value” hype that Futurism tried to give it.

also, I like that the transcript includes the seamless ad transition…which is of course for an AI product:

A quick word from our sponsor, Scale AI. Publicly available data is running out, so major labs like Meta and Google DeepMind and OpenAI all partner with Scale to push the boundaries of what’s possible. Through Scale’s data foundry, major labs get access to high-quality data to fuel post-training, including advanced reasoning capabilities.

As AI races forward, we must also strengthen human sovereignty. SCALE’s research team, SEAL, provides practical AI safety frameworks, evaluates frontier AI system safety via public leaderboards, and creates foundations for integrating advanced AI into society. Most recently, in collaboration with the Center for AI Safety, SCALE published Humanity’s Last Exam, a groundbreaking new AI benchmark for evaluating AI systems’ expert level knowledge and reasoning across a wide range of fields. If you’re an AI researcher or engineer and you want to learn more about how SCALE’s data foundry and research team can help you go beyond the current frontier of capabilities, go to scale.com slash Dwarkesh.

did these fucking dweebs seriously name their AI research team the “SEAL team”?

Bloomberg reports that “Humane’s team, including founders Imran Chaudhri and Bethany Bongiorno, will form a new division at HP to help integrate artificial intelligence into the company’s personal computers, printers and connected conference rooms,” per an HP executive.

congrats to HP on the launch of their new “you thought inkjet printers were shitty now? hold my aquifer and watch this” division.

but also:

HP is buying Humane’s CosmOS, bringing on Humane technical staff, and will get more than 300 patents and patent applications, Humane says in its press release.

this is a relatively cheap way for HP to set itself up as an AI patent troll and extract rent from other companies that are trying to do AI-related bullshit. (from 2017: Stupid Patent of the Month: HP Patents Reminder Messages)

Encryption lengths are getting long so you’d think it was high time.

that’s unrelated - AES-256 for example can be executed just fine on either a 32- or 64-bit machine. in theory there’s nothing stopping you from running it on an 8-bit or 16-bit CPU (although other considerations related to the size of AES’s lookup tables make this unlikely). from some random googling, here is an implementation of Chacha20, another 256-bit encryption algorithm, for 8-bit microcontrollers.

when we talk about 32 vs 64-bit CPUs, in general we’re only talking about the address space - the size of a pointer determines how much RAM the computer is able to use. 32-bit machines were typically limited to 4GB (though PAE helped kick that can down the road)

CPU registers can also be sized independently of the address space - for example AVX-512 CPUs have a register that is 512 bits wide even though the CPU is still “64-bit”.

What’s the relevance of either of those questions for an election that happened three months ago? I don’t like relitigating unspooled events.

my brother in Cthulhu, you started this post by saying:

this is where thinking Biden wasn’t doing enough has led.

you should decide if you’re for or against re-litigating things

Projecting your political beliefs and rationales on others is not Beeing Nice.

meanwhile, one paragraph above, you’re projecting an opinion onto me that I don’t have:

You’re welcome to your opinion that Biden or Harris would have been worse

I just find it almost comical that anyone thought Trump would be an improvement if not for the drastic outcomes we’re going to see.

OK…just to make sure I understand you correctly - the people you’re mad at, are people who either voted for Trump, or didn’t vote at all, because of their opinions about Biden’s response to the genocide in Gaza.

if that’s accurate, two questions:

a) what is your estimate for the size of that group of people?

b) how many actual individual people in that group can you identify by name? how many do you know personally? (vs having read a news story quoting them)

Explicit calls for ethnic cleansing were at least not on the table with the last administration.

Trump: we should ethnically cleanse Gaza

Democrats: 👏 HIRE 👏 MORE 👏 FEMALE 👏 GENOCIDE 👏 DENYING 👏 PRESS 👏 SECRETARIES 👏

I fear I’ve become something of an accelerationist in the past few days…

yeah, go ahead and pass this, you tech-illiterate xenophobic fucks.

we need to divide and conquer the fascist coalition. make them hate each other. make them consumed by infighting. give them more “oh I didn’t realize there would be negative consequences that affected me personally” moments.

there’s a whole lot of Silicon Valley techbro types who are on board with Musk and Trump because they think it’s all lower taxes, less regulations for their startups, and less “wokeness”. go ahead, pass a law that makes it a federal crime for them to click a GitHub download link. make it so that every Hacker News thread about AI is filled with American engineers bemoaning that they’re legally prohibited from keeping up with the state-of-the-art. make their startups uncompetitive because they’re required by law to pay inflated prices to subsidize OpenAI and other “American-made” plagiarism machines.

if anything Trump’s isolationist tendencies work in Palestinians’ favor, and otherwise he’s neutral compared to Biden/Harris.

I don’t even know where to begin peeling the onion of how incredibly stupid this statement is.

the genocide is (for now) over

“for now” is really doing some heavy lifting there

Netanyahu says Israel ‘reserves right to resume war’ and calls first phase a ‘temporary ceasefire’

but also, even if the ceasefire was permanent, that doesn’t mean the genocide stops.

if you bomb a bunch of hospitals, and then agree to a ceasefire, the hospitals don’t magically come back into existence.

ditto water treatment plants and similar infrastructure. Diseases spread in Gaza as sewage contaminates camps and coast

a ceasefire means that the “dropped a bomb on a refugee camp” aspect of the genocide stops. the “there’s a refugee camp with a never-ending stream of needless and preventable deaths” aspect of the genocide will continue unabated.