Google will invest as much as $40 billion in Anthropic
- Thread starter JournalBot
- Start date
Anyone getting the feeling that this is a repeat of "we lost money on every order but we make it up on volume" that was seen during the dot-com bubble? the only difference in the case of the dot-com bubble the companies were selling something people wanted (pet food delivered to your house) rather than something based around demand that is generated based on hype and baking that C-Suite types won't pick up a book to educate themselves (the sort of C-Suite types who rely on outside consultants who act as sales reps for these AI businesses).
Upvote
7
(8
/
-1)
Assuming Alphabet and MIcrosoft aren't complete idiots, my guess is that they're like 10% "AGI incoming" bet, and 90% "lose giant amounts of money, maybe become profitable longterm at some point,but suffocate all opposition whatever happens".
And so, almost everyone loses. Gargantuan wealth destruction, and monopolistic consolidation at the end. Feels like how evolution also produces inevitable catastrophic bottleneck conditions once in a while.
Upvote
6
(6
/
0)
Way around this:
First stop using their products and tell your friends!
For Search
Use Mojeek / Qwant / Ecosia / DuckduckGo
For E-mail, Cloud Storage and Office Use:
Proton, iCloud, Tutanota and others.
Then write to your politicians demanding they divest government funds from big tech hyperscalers and invest in local providers (or alt providers if you are American) like OVH, Hetzner, Leaseweb, Ventraip, Vultr, Digital Ocean, etc... tell them the economic damage they are causing and why these players are better. Demand they prioritise investments that are (1) local and (2) don't make business decisions that threaten their ability to continue to keep their workers employed.
If you are employed in tech as Bert Hubert says use your clout to make your higher ups do differently where possible.
To some extent this is now happening.
Upvote
3
(3
/
0)
Most people I know who use AI regularly are plonking down $20/month already. But that's not the problem. When the investor cash gusher stops, it's not going to be $1/month or $20/month, it's going to be at least $100/month. That's why Anthropic is pushing stuff out of the $20/month tier and into their $100/month tier. At $100/month, I'd bet almost all the casual users drop away. And I suspect given the size of investment spending, $100/month isn't going to be enough for AI companies to be profitable
Upvote
9
(9
/
0)
Apparently you can say anything in an Ars story about AI, and people will upvote it so long as it has the approved attitude, no matter how stupid it is.
Upvote
-6
(6
/
-12)
Estimations are difficult, because nobody exactly knows the actual costs of Tokens.
But according to Ed Zitron, a $200 Claude subscription lets you burn through $2000 to $4000 of Tokens.
I'm looking forward to the day when OpenAI and Anthropic have to be profitable. Then we'll see what people are willing to pay.
And I'm also looking forward to having actual LLM-generated software on sale with Token- based payments. Then we'll see how stable it is, how secure and what the support costs are.
As they say in Formula 1: When the flag drops, the bullshit stops.
Upvote
11
(11
/
0)
I guess we're devolved to the snake eating it's own tail portion of the AI bubble. I guess we all know what's next, it's just a matter of when and if any of us could figure that out we'd be very wealthy.
Upvote
-3
(0
/
-3)
I somehow signed up for API pricing when I installed Claude Code (while using claude.ai for free in a browser), and I get a bill for $10 every one or two sessions. That's less than $3/hour, extremely cheap for an agreeable, tireless, hyper-knowledgeable junior-plus developer.
I might balk if and when costs increase to ~$30/hour, but that's still cheap for a business or someone in a developed country trying to avoid being laid off.
Upvote
4
(4
/
0)
Yet we keep being told the CEOs driving this are such above average special unique snowflakes that every year we must pay them more money than most people will make in multiple lifetimes or else things will catastrophically fall apart.
Upvote
5
(5
/
0)
Except that jumping to only $30/hour seems optimistic. These companies are dumping billions and billions and billions into infrastructure and those costs have to be recovered via usage fees. If that $30/hour goes to $60/hour, now you're firmly in US labor rates for junior plus devs. And those junior plus devs are likely to be more useful than Claude. At least they have the chance to be able to work with minimal supervision.
Upvote
7
(8
/
-1)
Fools and their money.... chasing bad money with good...
Think about how many people could have been sheltered, fed, clothed, and educated with that kind of money. I know this is a sentiment older than I am, not a new thought at all, but try to see it fresh again, because it bears repeating.
Think about how many people could have been sheltered, fed, clothed, and educated with that kind of money. I know this is a sentiment older than I am, not a new thought at all, but try to see it fresh again, because it bears repeating.
Upvote
2
(2
/
0)
Now, in cash. The person you're replying to is more right than you are. Maybe it evaporates, but it's different than many other announced AI deals and that's what they were commenting on.
Upvote
1
(1
/
0)
The crucial point this article is missing is the development & emergent behavior of Mythos to uncannily detect zero-day exploits & chain them to execute an attack in seconds. It is a paradigm changing shift in cybersecurity equivalent to the development of the firearm in warfare in that it democratizes the destructive potential of any individual with access to the model. People will no longer require access to the resources & expertise of a nation state to execute sophisticated attacks on critical infrastructure.
All major US tech companies, the US & UK government, & soon major banks following emergency closed door meetings with the Treasury department are utilizing Mythos via Project Glasswing to actively detect vulnerabilities & write patches. Microsoft has officially confirmed it is embedding Mythos into its software development pipeline to detect these vulnerabilities prior to software reaching the market. Everyone is using Mythos to fight Mythos because an AI that is capable of detecting vulnerabilities & executing an attack in seconds needs to have a tool that detects those attacks, writes a patch & applies it in microseconds or fundamentally alters the input from the competing bot.
Unfortunately the genie is out of the bottle, so to speak, & the rapid deployment of AI vs AI is the only solution. The estimates for another model developing these emergent capabilities is, at most, a year out so the clock is ticking. It's an impending digital apocalypse, the Y2K of this generation & no one is connecting the dots or seriously reporting on this.
All major US tech companies, the US & UK government, & soon major banks following emergency closed door meetings with the Treasury department are utilizing Mythos via Project Glasswing to actively detect vulnerabilities & write patches. Microsoft has officially confirmed it is embedding Mythos into its software development pipeline to detect these vulnerabilities prior to software reaching the market. Everyone is using Mythos to fight Mythos because an AI that is capable of detecting vulnerabilities & executing an attack in seconds needs to have a tool that detects those attacks, writes a patch & applies it in microseconds or fundamentally alters the input from the competing bot.
Unfortunately the genie is out of the bottle, so to speak, & the rapid deployment of AI vs AI is the only solution. The estimates for another model developing these emergent capabilities is, at most, a year out so the clock is ticking. It's an impending digital apocalypse, the Y2K of this generation & no one is connecting the dots or seriously reporting on this.
Upvote
-12
(0
/
-12)
So they are dropping the mass customer base to chase the "premium" customers. Which is something a lot of companies have been doing over the past 10-15 years, and it is only accelerating...
Upvote
0
(0
/
0)
yeah if you believe the doom and gloom articles about Mythos's capabilites then it is a scary time in the cybersecurity world. That being said, I think you are overstating the potential ability of Mythos to find and execute 0Days. Like you see the big "Mythos found 230 0days!" statements, but when talking to experts, the overwhelming majority of the potential 0days it found were not actually exploitable because there were checks against those elsewhere in the code. Think of the AI model marking down a buffer overflow on a function as a 0Day, but the AI model doesn't account for the literal overflow error check the system does in a separate subroutine, so it really isn't a threat vector at all.
Here is an article about it from RedHat and their experience using AI models to surface vulnerabilities.
https://www.redhat.com/en/blog/navigating-mythos-haunted-world-platform-security
Upvote
4
(4
/
0)
Arguably its already happened, but the problem is that we have increased reasoning complexity so frontier models now burn 10x as many tokens as before and require 100x the compute to train.
For the economics to work you need the latter scaling to stop or its a perpetual capex sink arms race. But they cant stop or they will lose to competitors, its a catch 22 that requires continued money burning until the money is gone.
Upvote
2
(2
/
0)
If this is all true about Mythos (and I have doubts, but we'll see) then the endpoint isn't "give the good guys a faster Mythos to defend from the bad guys who also have Mythos" it's "the internet is no longer possible because an air gap is the only practical security." Because it's not just the one model, there's going to be more year over year.
Might actually be a good thing for global mental health, honestly.
Upvote
2
(2
/
0)
Anthropic is on a roll.
PocketOS founder blames ‘Cursor running Anthropic's flagship Claude Opus 4.6’ plus Railway’s infrastructure for data disaster.
The founder of PocketOS has penned a social media post to warn others about the “systemic failures” of flagship AI and digital services providers. Jer Crane was inspired to write a public response after an AI coding agent deleted his firm’s entire production database. The AI agent’s misdemeanors were then hugely amplified by a cloud infrastructure provider’s API wiping all backups after the main database was zapped. This tag team of digital trouble has wiped out months of consumer data essential to the firm’s, and its customers, businesses.
article at:
https://www.tomshardware.com/tech-i...-tool-powered-by-anthropics-claude-goes-rogue
PocketOS founder blames ‘Cursor running Anthropic's flagship Claude Opus 4.6’ plus Railway’s infrastructure for data disaster.
The founder of PocketOS has penned a social media post to warn others about the “systemic failures” of flagship AI and digital services providers. Jer Crane was inspired to write a public response after an AI coding agent deleted his firm’s entire production database. The AI agent’s misdemeanors were then hugely amplified by a cloud infrastructure provider’s API wiping all backups after the main database was zapped. This tag team of digital trouble has wiped out months of consumer data essential to the firm’s, and its customers, businesses.
article at:
https://www.tomshardware.com/tech-i...-tool-powered-by-anthropics-claude-goes-rogue
Upvote
3
(3
/
0)
I think AI is rising to "rogue engineer" status. The nefarious actor you can blame for all your incompetences.
So many things went wrong there, only one of them was blindly trusting the AI. And that was the least interesting one. Why does anything have the access needed to delete the production database and wipe all backups? That's legit insane. They were always just one oopsie away from disaster.
So many things went wrong there, only one of them was blindly trusting the AI. And that was the least interesting one. Why does anything have the access needed to delete the production database and wipe all backups? That's legit insane. They were always just one oopsie away from disaster.
Upvote
3
(3
/
0)
I love the way he doesn't question his own process choices.
"I put all the eggs crucial to my business into a single basket and assigned a non-intelligent piece of software to sit on them."
"Who knew this could end badly? Who knew?"
It's his competency that's most in question.
edit: You never see the ninja ... until it's too late.
Upvote
2
(2
/
0)
Except a junior programmer will learn, and hopefully can be trusted to work without close supervision after a while. That's not to say an AI coding assistant doesn't have value, but it's not a fair comparison.
Upvote
3
(3
/
0)