These people are religious fanatics, and should be treated as such. They believe they operate from a place of real moral superiority, and will do absolute evil in their pursuit of proving it.
11,569 karma · joined May 23, 2017
These people are religious fanatics, and should be treated as such. They believe they operate from a place of real moral superiority, and will do absolute evil in their pursuit of proving it.
Nothing like
- Major bank consolidation
- Major media condolidation
- Major tech consolidation
- Major grocery store consolidation
etc... to really provide wonderful competitive options.
Hence - the man is much more dispensable...
---
The water in my water bottle is much more dispensable than the toothpaste in my toothpaste tube. But both can be dispensed (in both meanings - distributed or replaced).
My experience (both in and out of management, and as a very high level IC at a mix of companies) is that while the management layers generally didn't understand what was going on in the code - the decent managers tended to have a "gut check" person they trusted to go and ask technical questions, and bounce ideas around with.
This was their "expert" who they trusted to be close enough to ground truth to tell them if a thing will work, how hard it will be, if there are shortcuts, etc... Often times, it wasn't even strictly technical questions that would get answered, it will be things like "how much of this project is technically in my sphere of control, and how much do I need to go align other teams? How many teams? Which teams? Which projects? etc..."
My problem is that AI SUCKS at this job. So yes - we're getting pushed into management roles as ICs... but instead of a decent subject matter expert who still understands ground truth... you get an LLM blabbering on about seams and load-bearing joints, and clickbait style headline summaries, but completely non-nonsensical judgement calls about how to interact with the people who make up the company.
Ultimately, work is also politics - and it's not just "writing code" that matters. It's knowing what code is worth writing, and when you should write it. And an AI will happily tell you that a single change-set in a single repo is "substantially larger then expected" and put a timeline of "two weeks" on the completion (especially if anything Jira related happens to be in context... blegh), and then knock it out in 5 minutes time... While also spouting off about how easily it can add "new feature X" it just needs to get PRs in across 5 different repos, owned by different parts of the org with non-aligned managers.
Not to mention, loud as all get out at that size.
It also needs a decent amount of maintenance to keep it running (or it needs to run frequently).
Personally... Ev with a swappable battery seems like a much better solution.
The only reason this even sorta makes sense right now is that we're in a really weird transition period where gas is available everywhere and it's still hard to find batteries/chargers.
This isn't a tech problem, it's an infrastructure problem. It should be treated that way.
Personally, I'd also like to see party influence removed from elections. Completely removed at the time of actual election (ex - absolutely not allowed to list party on the ballot) but also reduced at the primary level (ex - no more closed primaries, maybe eliminate partisan primaries entirely).
Politics isn't a team sport with winners and losers, and this attitude (especially concerning given the recent moves by republican leaders to explicitly disenfranchise and withhold support for locations that trend democratic) is utterly reprehensible (which is fairly in theme with their overall moral character).
We're one nation. Politicians serve the populace, regardless of whether those people voted for them or not.
Further
> If you believe capabilities will go to "Virtuoso AGI" or beyond, then it's not enough. A smart enough agent can appear to be loyal, transparent, etc. but how would you know? If your bank balance keeps going up 20% YoY, is the agent optimizing your long-term flourishing, or preparing for a rug-pull?
> Now, if you could somehow white-box these LLMs and mechanistically _prove_ that they were acting as your fiduciary, then that would get us somewhere.
This is exactly the same problem we have today with those who are bound by these rules (humans - to be clear).
The idea is not that it's impossible to violate these rules. It's that these rules create a boundary for expectations in the relationship, with legal teeth.
Ex - If I want an LLM that puts together a shopping list for me, with links to buy online... I expect that LLM to be serving my interests. If a provider (either inference or model weights) wants to influence the choices that LLM makes because they make backroom deals with specific store - I'd like that to be illegal.
Same for competition
Ex - If I want an LLM to put together a product that competes with the provider of that LLM (either inference or model weights) and that LLM refuses - I'd like that to be illegal.
The idea is not that they can't possibly do those things. The idea is that we preemptively define relationship expectations, and set hard boundaries around what things we fine/punish.
Misaligned models are a problem everyone wants to solve. Models created by misaligned companies are a god-damn disaster.
They favor keeping the system that is producing rampant inequity stable. With a laissez faire approach to social issues.
As opposed to republicans, who favor socially regressive and oppressive policies, double speak, and who's mantra is mostly "Don't get caught. But if you do get caught - deny, deflect, denounce" rather than anything resembling ethical behavior.
Neither of them are particularly good options for governance. They have no motivations to "learn" anything.
One is a fast moving, malignant cancer - the other a slow moving auto-immune disease. Both kill the host.
We desperately need fresh air.
What you really want is fiduciary duty - A fiduciary is a person or organization that is legally and ethically bound to act in the best interest of another party (think financial advisor, attorney, guardian, trustees, etc...)
And I cannot agree more. I think we should be shooting to enshrine required fiduciary duty into law for LLM providers as quickly as possible.
To recap why:
Legally, fiduciary duty means basically 4 major tenets must hold
1. Duty of loyalty - it must put the interests of the client ahead of their own
2. Duty of care - it must make well-informed, prudent decisions
3. Avoidance of conflicts - it must avoid situations where personal gain conflicts with client obligations
4. Transparency - it must disclose fees, risks, and conflicts as soon as possible
---
You can't have a reliable "agent" if those things aren't true, because an agent is (by definition) someone who is working on your behalf, for your goals. If it's not working on your behalf, for your goals... it's not your agent, it's an opportunistic spy (double agent) waiting for the best moment to sell you out.
Ex - go check amazon, they're all the "dev kit" versions of ESP32.
Hell - even Alibaba has an overwhelming majority of the kits in the SBC form, rather than the straight MCU (although they at least exist in that form there, unlike amazon).
Even on digikey... there are like 4 listing for the straight MCU, and hundreds of listings for PCB board versions.
So I think for most consumers purchasing these... functionally they end up buying a form factor that is much closer to SBC than MCU.
---
From my perspective... I'm old enough to remember when Arduinos were the hot shit in this space, and now ESP32 is basically the same value prop at 1/10 the price point (or better, if you are actually willing to solder it to the PCB yourself... then you get down to 1/50th the price point).
Instead I deal with a 30+ repo clusterfuck (technically we have 60+ services, but I only have to run half...) that is held together by hopes and prayers, takes literal hours of actual effort to bring everything up to date on master, and has become a fractured hellscape where people are afraid to leave their tightly constrained silos of service combinations.
Long story short... I will take a bad monorepo over bad multirepo any day of the week.
It just puts Google in the same sandbox Google puts everyone else.
Typing this on a pixel running graphene.
1: I think people should be able to replace chips on an RPi as they desire, so I don't love that they're actively preventing this
2: As a buyer, I absolutely want to know if I'm buying an RPi that has had chips replaced
3: A shady seller changing the SKU to mimic a better board after changing chips is just flat out fraud.
---
It's the classic "Mattress tag" case - it should be illegal to remove as a seller, and fine as the final customer.
But that's really hard to functionally enforce.
I'd argue the best possible model for goods that have no marginal costs (aka IP) is the "pay me to create it" model.
As in, let a creator decide how much they value creating the work. Then pay them that to create it - then release it free for everyone.
---
No ridiculous societal restraints that cost time and money to enforce (poorly) as IP law. Compensation is happening. Creating is encouraged.
Practical constraints make it unlikely we'll get much closer to this than Kickstarter (and equivalents) but it's a very honest model, that works with intellectual property by default.
here's a decent recap (now a couple years out of date)
And it's hardly like we're "super concerned about the tests because of the test itself" here... given we've triggered nearly 1000 underground nuclear blasts in the US alone, across several states (at least 6).
So... yeah - a relatively unstable state gaining nuclear capabilities sucks, but the interesting part of the discussion in the article is why we're seeing activity years after the testing.
They use bank like Ally or Discover with no physical branches.
They use a mortgage provider like Rocket mortgage with no physical branches.
They use a medication delivery service with no physical customer facing pharmacies.
They have an employer that only facilitates reimbursement for expenses via online tools.
etc...
I guess "survive" has a sliding scale, but if I lost access to critical accounts... my life is going to FUCKING SUCK in a non-trivial and very impactful way almost immediately, on many fronts.
And if your answer to that problem is "well, just call them"... then we're right back to the point the article is making: "An account’s security is still dictated by the weakest recovery method"
Passkeys aren't a meaningful improvement in security - assuming you do actually have decent password hygiene like a password manager.
https://developer.chrome.com/docs/ai/built-in/overview?_gl=1...
Some are quite good at limiting connections, monitoring devices, and contacting students for suspicious traffic.
Functionally, it's a "workflow" runner (ex - you can mostly treat it like a queue, where you've got workers that are picking up work to do).
But it wraps a couple of pretty handy features on top like:
- It preserves most arguments to actions, and it makes system details deterministic for retries (ex you can re-run a workflow at a later time, and temporal will make sure the code sees details like date and time as though it were the original run, and will yell at you if you try to write code that won't be deterministic on retries)
- It supports very long waits easily. (ex - very easy to have a workflow do a couple things, wait a week, then do some more things).
- It has decent profiling and UI tools
- It can "restart" a failed workflow deterministically from the step at which it failed (using details from the original point)
---
Basically - it's a background worker service that's put a lot of time and thought into ways to handle failures better.
It absolutely still has some considerable pain points though, and I find it difficult to use for larger tasks (ex - their message gRPC size limit of 4mb is a b*&^% to work around, since it often breaks a lot of the utility they provide, and the history cap at 50mb is also really painful in certain situations.)
Really - I think it was just the right tool at the right time to make calling LLMs with long waits relatively easy and somewhat foolproof.
The modern structure of society has effectively given government a complete monopoly on force and violence.
So the second a disenfranchised person responds with violence, it's easy to spin it as their fault, treat them as an instigator, and then respond with overwhelming force with police.
Regardless of your opinions on morality, the Luigi Mangione saga is literally this playbook.
And really, it only gets the attention it does because a majority of people literally despise their insurance providers in the US.
Modern feudalism will be the right to control your access to "required services" needed to be considered a normal member of society.
Ex - who needs to get physically violent with you when locking you out of the banking system, removing your ability to speak in online spaces, and crippling your ability to use network based services is much more effective at disenfranchising you, without the bad publicity of direct violence.
You won't be harmed physically, you'll just be permanently disenfranchised and second class.
So... No. Even if he means what he's saying, it's irresponsible of us not to consider his possible motivations.
I deploy browser extensions to millions of devices, and small local models that I can run in that space are incredibly appealing.
But the "per user" model is just especially difficult to justify here, and comes with a TON of fucking downsides like usage tracking, feature availability, and product structure (ex - is a trial user a user? how long? etc...)
---
And it's not even like they're justifying it with model download support or caching... they shove it entirely onto hugging face, or have you self serve.
As their literal target audience - I'd rather buy the weights for a fixed price (even a high one), and not let them influence my product design decisions with saas style pricing.
It's a fairly generic body shape, but the headlamp and grille shape is distinctively "happy" instead of "aggressive" (this is a real thing, and it's intentionally designed for in lamp/grille shapes on cars)
Personally, I don't want "aggressive" near my cars. They kill enough people already, and I'm tired by how common it is for manufacturers to push aggressive body shapes.
As in - I make software for a living and they can fuck off. Produce the item. Sell the item at a cost you find reasonable. Convince me to buy it again with an upgrade that's worth buying.
Period.
There's no "service" happening here. If anything, the proposed cost model just makes me incredibly unlikely to engage at all, because it means they're going to insist on trying to track my users (a thing I don't like doing...) with whatever bloatware they shove into their "sdk".
If anything - I actually think we're at a spot where the mask is slipping a bit, because the US was slow to adapt its media distortion to the internet, and the financial interests in the US are stretching even farther away from general sentiment.
So the notice is because they've fumbled, and now are beating the drum extra hard to try to make up ground.
Uber doesn't buy the cars because it's not profitable.
They let the driver eat the depreciation, assume the capital risk, and deal with lemons and accidents.
No one drives Ubers thinking they're going to get rich, and I sure don't see companies buying up vehicle fleets to put into service on Uber.
Because it's a money losing proposition.
Why do you think Tesla is trying to also offload it onto suckers?