startupsfail
startupsfail on Hacker News
This potentially can be a supply chain attack at a massive scale.
When you do not have a seat at the table, you are not in the game, and winning a game is an impossibility. As long as you are a player, it is remains an option, if perhaps not win it somehow, but at least drag it to a draw, or change the rules,
or make a loss to be survivable.
It can be a win-win. Simply having a seat at the table can be a win.
The system in question is a distributed system, an interaction within that system such as "confession" involves ridiculous amounts of distributed processing, far beyond two nodes that were participating in that original exchange.
"Wero lets users send money using just a phone number — no IBAN, no card, no intermediary."
As long as all the other cards still get acceptance, this seems like a great system.
Right way of resisting is not to quit your job, but to observe, be friendly, and gradually push away toxic people.
Claude is already known for its attempts to send emails to FBI ;)
The results are repeatable. Models are performing with predictable error rates on the tasks that these models had been trained and tested.
The default could be that a background upgrade should not be a foreground stress test.
Imagine you are driving a car and from time ro time, without any warning, it suddenly starts accelerating and decelerating aggressively. Your powertrain, engine, breaks are getting tear and wear, oh and at random that car also spins out and rolls, killing everyone inside (data loss).
This is roughly how current unattended upgrades work.
Is there a good reason why upgrades need to stress-test the whole system? Can't they go slowly, throttling resource usage to background levels?
They involve heavy CPU use, stress the whole system completely unnecessary, the system easily sees the highest temperature the device had ever seen during these stress tests. If during that strain something fails or gets corrupted, it's a system-level corruption...
Incidentally, Linux kernel upgrades are not better. During DKMS updates the CPU load skyrockets and then a reboot is always sketchy. There's no guarantee that something would not go wrong, a secure boot issue after a kernel upgrade in particular could be a nightmare.
I'm curious, where it has to be disclosed? Like if a company would pay a few legitimate reddit account owners to review their post and upvote, and would disclose this activity in the DISCLOSURES.txt available on their website, would that be legal?
Where would one find some reddit users willing to do such reviews, by the way?
There are still blatant failure modes, when models engage into clear sycophancy, rather than expressing enthusiasm, etc.
I'd guess, in practice a benchmark (like this vibesbench), that could help catching unhelpful and blatant sycophancy fails may help.
There is a nice essay from Paul Graham that starts with:
> The word "prig" isn't very common now, but if you look up the definition, it will sound familiar.
It's not a question of mimicking, it is interesting what is current within the teenage/student community. Adult population runs out of steam at some point.
I'm curious, is there some meaningful way for geriatric millennials to use Tik Tok?
Without being sucked in into doomscrolling and content consumption? Produce content? I'd guess it should be possible to play with the thing somehow...
Seems like an interesting story, Ashawna - she was about 25 at the time, and as per Wikipedia, already worked on the military projects - the Sprint Missile System, and was at Xerox.
> The processor was reverse-engineered by Ashawna Hailey, Kim Hailey and Jay Kumar. The Haileys photographed a pre-production sample Intel 8080 on their last day in Xerox, and developed a schematic and logic diagrams from the ~400 images.
Yes, executioners do proliferate this way. They tend to run out of murderers quickly though, then use any other excuses to execute.
Wow, I'm looking at current "Open Shuttles", a license to use 4KB of SRAM in the project is $2500. But it comes with Wishbone Bus interface!
> 1024x32 Commercial SRAM
> CF_SRAM_1024x32
> Commercial SRAM: 1024 words x > 32 bits (4KB) with Wishbone Bus interface
> Area: 0.17mm²
> GPIOs: 0
> License: Commercial - $2500 per project
Junior engineers now learn from AIs. And AIs now learn from RL cost functions. And RL cost functions are being set by PhDs, with little to no production grade engineering experience ;)
The result is interesting. First, juniors are miserable. What used to be a good experience coding and debugging, in a state of flow is now anxiously waiting if an AI could do it or not.
And senior devs are also miserable, getting apprentices used to be fun and profit, working with someone young is uplifting, and now it is gone.
The code quality is going down, Zen cycle interrupted, with the RL cost functions now at the top.
The only ones who are happy are hapless PhDs ;$
There are enormous microcode, firmware and drivers blobs everywhere on any pathway. Even with very privileged access of someone at Intel or NVIDIA, ability to have a reasonable level of deterministic control of systems that involve CPU/GPU/LAN were long gone, almost for a decade now.
The same argument was there about needing to be an expert programmer in assembly language to use C, and then same for C and Python, and then Python and CUDA, and then Theano/Tensorflow/Pytorch.
And yet here we are, able to talk to a computer, that writes Pytorch code that orchestrates the complexity below it. And even talks back coherently sometimes.
> But as estrogen levels shift in perimenopause and beyond, this intense drive to please and nurture others begins to diminish. What replaces it isn’t bitterness. It’s clarity.
It's not clear how anxiety, mood swings, brain fog, inability to remember faces, fear, aggression are somehow being called "clarity".
<think>Ok, the user is claiming that... </think>
....
They are likely making informed decisions. Disney/Pixar are players, not bystanders.
Can you give an example of some part of the physical world or infosphere that an LLM can't model, at least approximately?
It's funny how Men vs Women in Science, or IQ standard deviation differences still result in such heated debates, while at the same time...
Below is the worst quote... It is plain wrong to see an LLM as a bags of words. LLMs pre-trained on large datasets of text are
world models. LLMs post-trained with RL are
RL-agents that use these modeling capabilities.
> We are in dire need of a better metaphor. Here’s my suggestion: instead of seeing AI as a sort of silicon homunculus, we should see it as a bag of words.
There are a lot of people around who are playing victims to get benefits. And relentlessly pushing their agenda.