Especially since it decomposes the image into a semantic vector space rather than the actual grid of pixels. Once the image is transformed into patch embeddings all sense of pixels is entirely destroyed. The author demonstrates a profound lack of understanding for how multimodal LLMs function that a simple query of one would elucidate immediately.
The right way to handle this is not to build it grids and whatnot, which all get blown away by the embedding encoding but to instruct it to build image processing tools of its own and to mandate their use in constructing the coordinates required and computing the eccentricity of the pattern etc in code and language space. Doing it this way you can even get it to write assertive tests comparing the original layout to the final among various image processing metrics. This would assuredly work better, take far less time, be more stable on iteration, and fits neatly into how a multimodal agentic programming tool actually functions.
Yeah, this is exactly what I was thinking. LLMs don't have precise geometrical reasoning from images. Having an intuition of how the models work is actually.a defining skill in "prompt engineering"
Asking your favorite LLM actually helps a lot. They generally are well trained on LLM papers unsurprisingly. In this case though it’s important to realize the LLM is incapable of seeing or hearing or reading. Everything has to be transformed into a vector space. Images are generally cut into patches (like 16x16) which are themselves transformed by several neural networks to convert them into a semantic space represented by the models parameters.
But this isn’t hugely different than your vision. You don’t see the pixel grid either. You have to use tools to measure things. You have the ability over time to iteratively interact with the image by perhaps counting grid lines but the LLM does not - it’s a one shot inference against this highly transformed image. They’ve gotten better at complex visual tasks including types of counting, but it’s not able to examine the image in any analytical way or even in its original representation. It’s just not possible.
It can however make tools that can. It’s very good at working with PIL and other image processing libraries or even writing image processing code de novo, and then using those to ground itself. Likewise it can not do math, but it can write a calculator that can do highly complex mathematics on its behalf.
I’d note they’re not mutually exclusive revenue streams and both add meaningfully to their value. I think the reality is they peaked the first one and growth is in the second one. Subscriptions that are sticky however are much more valuable individually than an advertising tier user. But if you can cater to both and not downgrade subscriptions to ads tier you win in two parallel markets via the same platform. This is not a bad business strategy. But they need to not lose the subscriptions and their reason for being in the quest for growth or they’ll see nominal growth with decline in value.
note: I hate ads so I'm not trying to manifest this, but can you explain why you're so sure of this?
To me, it seems like they "should" (for greed reasons, I mean, not for my happiness) hike the prices of subscriptions aggressively while keeping the ad-tier attractively-priced, moving as many people as possible over. This increases ad revenue and allows more YoY growth if their ML can manipulate you into more watch hours in 2027 than you do in 2026.
Sure, some people like me will probably drop Netflix before they'll pay $35 a month or endure ads. But the current delta is only $10. I suspect they can make $10 a head in ad revenue in a year -- and if they can make $15, they would break even if they lost 3 ad-free subscribers but gained 2 back onto the ad tier. Anything better than those numbers would be a net gain.
Because the way subscription revenue is accounted for is a present value of the expected duration of the subscription and ad revenue is cyclic and varies throughout the year across individuals, cohorts, and the population. They’re also generally different markets - people willing to endure ads are either unable to afford the subscription or cheap, in either case it’s not unreasonable to expect the impression value for ads is pretty low.
Welcome to 2022 or so. I thinl Hulu started it, but yea. Many "premium" services are back to ads again on the lowest tier. It's probable more expensive than the highest tier 10 years ago as well.
AKA - it’s the department of defense in the same way Robert Kennedy was named Robert but went by Bobby sometimes. Trump doesn’t get to change the name, just assign an unofficial nickname that he thinks sounds more tough. Sort of like his pretend tough guy Secretary of defense that dresses up as a Secretary of war for TV moments. The fact they drag along the entire military and its leadership in their charade is embarrassing, and the asinine nickname is expensive and likely causes operational confusion.
Why is alignment necessary? In our system compromise is the typical alignment sought where no single view dominates the decisions or direction. With enforced alignment no compromise is more than not necessary it’s not possible. That’s the dysfunction of the present because there’s a perception that holding office entails enforcing alignment, and opposing voices not only need not be heard but are forcefully silenced. However the system we have in the US doesn’t allow for that, and explicably, it’s even more dysfunctional than normal. Sooner or later they have to stop and compromise, over throw the system, or be removed. That’s precisely how it’s designed to work.
So, you shouldn’t be silenced, your opinions should be heard, and to the extent they’re reasonable, they should be considered proportional to your ability to influence. The more to which this is prevented or ignored the more unstable the system is.
There’s a difference between forced alignment and rejection of falsehoods. On each of these the response to the questions were investigations in the public space, especially vaccines, but through various processes. This is the opposite of forced alignment - this is deliberately considering opposing views. That doesn’t mean a decision isn’t eventually reached, and the fact we are relitigating all of these year over year even on verifiable facts shows there is no forcing of alignment by the system. I think people don’t realize what forced alignment looks like - that’s Soviet Russia, Maoist China, Xi China, North Korea, etc, where dissent is not tolerated, not allowed, and alignment is forced.
Well, HN isn’t a part of the system I refer to, the system is the constitutional republic under the bill of rights. And you’ll notice as forced alignment is attempted in our system it gets litigated and often fails. The policy system is generally deliberative requiring input and consideration at minimum. Everything is reviewed by many layers of judicial review for fairness, and every six years the entire executive and legislative branch could undergo a total revolution. What i think is false is considered but isn’t by mandate so, no matter who I am - president, senator, HN poster. What is considered true today can be repudiated tomorrow.
These “commitments” aren’t generally contractual obligations beyond some clawback or penalty for not fulfilling them. Don’t forget Microsoft and oracle retain the infrastructure built. It’s generally reserved capacity with minimum spend and a maximum spend as the quoted $1.5T. The spend also typically ramps and there are triggers and out clauses built in. These aren’t bonds or debt or what they imply as “coming due” in the article.
People keep drawing parallels to the housing crisis and ensuing financial crisis. Those were hundreds of millions of individual contracts spread across thousands of banks and debt holders. There was little understanding at any individual institution of the correlation risks with subprime or the systemic counterparty risk (I know I sat in the center of core risk at one of the big systemic banks at the time). In this the institutions are all large, know their counterparts well, know OpenAI, etc. Despite all outward appearances none of them are signing up for insolvency, and none expect a bailout, and none WANT a bailout. It’s not good for any CFO’s career to be the one forced to give up the companies ownership to the US government and they at minimum will avoid this at all costs.
I’m not saying this isn’t all a giant bubble, but it’s more akin to the dot com bubble which was more of a shakeout than long term consolidation. It’s hard to say the www isn’t an intensively valuable set of infrastructure and services even if pets.com went out of business (and its business model has turned out to be pretty viable since).
I suspect this government isn’t receptive to commentary from anyone other than only one person. While I’d never discourage anyone from advocating their beliefs this feels like at best a waste of energy. They are going to do it because they decided to do it - the solicitation of comments is performative and required. The only way to stop it is via the courts and by voting next November.
There's even precedent for the current president's agencies compiling some pretty sketchy "comments" in the past due to not doing basic sanity checks on pretty obvious fake comments that happened to support their agenda, like when supposedly seeking input from the public about repealing net neutrality[1]. There were so many duplicates that only thirty 30 unique comments made up 57% of the overall total, and the second most common "name" among the authors was literally "The Internet".
No one in the current administration cares about what random members of the public think about their policies, and that's by design. Even the government positions that are intended to be permanent across administrations aren't a safe bet at this point with was things have been going
You are 'this government's best friend, advocating for their opponents to give up and quit. In a remarkable pattern that I never thought I'd see in the rugged individualistic, idealistic, freedom-loving USA, a large group is literally self-defeating: They defeat themselves before even getting out of bed.
That's why your opponents are unstoppable - because you don't stop them. The performative nonsense is their aggression display.
They still want to win the election. Political and policy outcomes aren't all or nothing; the more they see, the more it will nudge them in whatever direction you want. Others will see it and it will nudge them too. If one person didn't embrace being a quitter, others would do the same.
> the root problem is not incompetence, it's that half of America wanted exactly this, for a second time now.
That is the same psychology I described in the GP: Instead of looking in the mirror and figuring out what they need to do better, they blame outside forces. It's victim psychology - powerless, someone else's fault, etc.
Your group failed; people didn't vote for it because you are well-known quitters and whiners and victims - and losers; you're ok with losing and quit when it happens - and you conduct shitty politics as a result. Who votes for that? Who even can stand to listen to it - it's sickening, depressing, disheartening.
The right wing says, 'we believe in X and we won't be stopped no matter what; we will never give up'. That gets votes. That gets things done.
Get out of bed, stop crying, and get to work. That you still hold on politically with this victim psychology shows how bad the right wing's message is. Never give up, never even talk about it.
> Instead of looking in the mirror and figuring out what they need to do better, they blame outside forces.
I'm not American, so let me ask you: what do you think they should have done better? Maybe a not presenting a woman candidate? I ask because that's the most common denominator in their last two election losses.
Take a good position - the GOP leaves the field wide open, maybe as simple as 'Liberty and Justice for All' (part of the American pledge of allegiance that everyone knows) - and fight for it, stand up for yourself, against all enemies and attacks, and fight until you win. Take punches, never go down, keep going, have a winning strategy and keep going until it works. Nobody will believe in you until you do it.
The Dems never confront the GOP and always fold. I think if they fought, successfully, for almost any policy, they'd attract a lot of support.
> the most common denominator
Not at all. How about they were Democrats, Senators, .... If it stand out the most to you, maybe ask why?
Taking performative actions you are certain will fail unnoticed is a waste of energy. I most certainly didn’t advise doing nothing, this seems like a hyperbolic take that ignores what I suggested was more impactful. By extension, encouraging others to take those actions is productive and there are other actions I didn’t enumerate that are productive - I didn’t intend to be exhaustive in all actions that could be productive, just that this specific action of commenting on their preordained policy decisions is pointless. I don’t see any argument here that in any way refutes that so I assume you agree.
> Taking performative actions you are certain will fail unnoticed is a waste of energy.
It's true, but usually in the opposite way you intend. If you go into ventures thinking they will "fail unnoticed", you certainly will fail. For example, who would hire someone or invest in someone with this attitude?
If you go in determined to succeed no matter what, there are no guarantees but you have a good chance.
Comments certainly contribute - the only risk to their power is people like you mocking them. And have you ever seen a successful team where some people mock others doing work?
I’d note you’ve not really proposed anything related to how commenting on policy in this administration achieves anything just pointed to cases where negative thinking brings bad mojo or something. Along that vein, I have been trying to grow wings and fly to the moon for the last 15 years but despite the investment of effort have not succeeded. It’s probably because people kept telling me it’s not possible and a waste of time, by the reasoning here. If I could get people to stop mocking me when I flop off the roof onto the ground I surely would succeed?
There are activities that are absolutely not worth doing because their chance of succeeding is zero percent even with the strongest of desire for it to not be zero.
I posit because this administration literally does not care what you have to say only what one individual on this planet has to say there’s no point in trying to reason with them. The only action that will work to stop them is in court and in the voting booth, and by proxy activities that magnify the value of either of those activities by getting others to participate. I think protest is a much more effective use of energy for that reason - it energizes like minded people and when this administration reacts with brutality to opposing opinions it shocks people not aligned with dictatorial oppression - which is almost everyone. But participating in the comment processes of their regulatory capture? Waste of time with zero chance of causing even an iota of change no matter how hard you wish it to be otherwise or how much you ignore the naysayers. Spend that energy growing wings and flying to the moon.
The government isn't one person, and I think both bureaucrats and judges are actually quite receptive to lots of people - only it's nebulous to who and why. Trying to please, and hoping to get rewarded, but neither you or they themselves are 100% certain of by who. Opaque power structures, everyone's paranoid, including the powerful.
Actually the comment collection process has in fact caused changes in policy over many administrations, largely because the policy makers were interested in achieving some goal aligned to their function and believed in our system. When commentary went strongly against their policy there was often a step back and reassessment. This administration appears to have two goals: maximally hurt people who hurt Trumps feelings and create the deep state they lamented but never really existed until now to continue to maximally hurt people similar to those who hurt trumps feels in perpetuity.
The right way to handle this is not to build it grids and whatnot, which all get blown away by the embedding encoding but to instruct it to build image processing tools of its own and to mandate their use in constructing the coordinates required and computing the eccentricity of the pattern etc in code and language space. Doing it this way you can even get it to write assertive tests comparing the original layout to the final among various image processing metrics. This would assuredly work better, take far less time, be more stable on iteration, and fits neatly into how a multimodal agentic programming tool actually functions.
reply