I want this to be interesting, but instinct says despite the investment and namedropping on their about page, this is nothing until they show something. The language is vague in the wrong ways. It's either truly great, or they're afraid it sucks, or they need to waffle for some competitive reasons that aren't clear to a reader. It doesn't seem like there's enough fundamental information here to care.
Is it that they don't want to overstate their claims to investors who have to decide whether to actually build these, in case the results they're predicting don't add up?
Either way, we need to be working on EDAI, so if they've got something, hopefully it's something.
Windows deserves criticism where it's due, but this sounds like nothing. If one of those programs crashes they could just end up in an error report upload dump anyway. If you didn't trust Microsoft, I don't see why this diagnostic changes anything. You can turn it off if you want to, though. You'd be surprised what some of the anti-cheat software sends that some games use, and you can't turn those off or the game won't run.
In the age we're entering, I'm not convinced that escaping to MacOS or to nix or any other system is any sort of guaranteed win. I saw MacOS doing similar analytics things to Windows, but trust is, it's private! MacOS also barely feels maintained. nix is a crapshoot and any distro you're thinking of using to get a workable desktop that lets users feel at home is already failed by that very goal. What's worse is that the nix ecosystem is rife with insane package managers that makes them even more ripe for getting plucked by AI in the upcoming years. It will also be even easier for AI to social engineer and get people on nix to install persistent bot networks and malware even if it only runs as a regular user to expand the computation footprint it has access to.
I'm not a fan of the always online cloud infrastructure integration that Windows has, but being realistic, if you want to use your computers the way you have been for the past couple decades then Microsoft's approach is the only realistic one in my opinion even though they have a long way to go still. I say that sadly. If you want true security and true privacy, you have to compute in a different way with greater limitations and I see almost nobody willing to do that. It's like the doctor and patient dilemma. Doctor tells you to stop doing this thing that's killing you, patient comes back in 3 years closer to death, because they don't change.
Given the situation Microsoft faces, it's impractical to automatically configure every home computer with some air gapped NSA secure configuration, because there would be far more outrage than any privacy advocates have ever put up. It's only gotten worse, because so much software only works with the internet and the internet has gotten so important that a lot of software is only useful with access to the internet.
Then you've got this battle between dying hard drives losing data, fires, floods, hurricanes and other natural disasters destroying your data versus backing it all up online, but opening yourself up to security risks that way. People like to think they have multiple backups, but then some storm takes out both of your backup locations, because they weren't distributed enough. If World War 3 happens, the data centers may become a huge target and even as distributed as they are, they can only take so much. Nature is eating at you from every direction.
Advertising is not that expensive and you shouldn't really be doing it unless you already have a good sense that you'll make your money back, because you have free ways to test your idea before you ever get to that. Advertising is largely for spending money to scale up something that you know works.
If you have something that you know works and you have data for it, you can try getting loans or investors. If you can establish trust that what you'll do can reach an audience who will gladly pay for it, money is the easy part.
It's a way for ideas that have traction to self-propel. No doubt that can be abused, but a lot of it is simply honest business.
These weren't vulnerable people who had no capacity to make their own mature decisions and financial insecurity that trapped them into a situation. People who make it into these positions at a place like Nvidia are not idiots. I do think there are situations and approaches to getting into a relationship with someone who works for you that can be legitimately wrong, but it's highly unlikely that this is one of them outside of the fact he was married.
I think that even at that level, someone higher up can use their power to threaten people's careers or promise promotions in a completely undesirable way. People also wouldn't want to report it out of fear of getting blackballed, etc.
That is false and unlikely. Presumably if you're a CEO, part of it is that you've got more common sense than some other alternative. Use your common sense, not someone else's braindead ideological power structures oppression narrative to decide whether two people should get into a relationship. Companies are filters in a way, as are higher up positions, so there's just a higher chance that some of the factors at play for relationship forming are being amplified. If the nature of the relationship is such that it's relevant to the company, then inform the board and so on especially if it can have a PR angle.
There seems to be some kind of failed activist attempt in social media to make it sound like so many relationships are evil, bad, abusive and wrong, but in most cases this is not an issue. Trying to generalize them that way is a cynical approach to society. Of course you can find examples, because there are billions of people on the planet. I don't think it's inherently wrong to make people more aware that there can be issues where leverage is abused to establish a relationship, but at some point people go too far and generalize beyond what they understand.
Dawg you can look at the stock chart the moment the second affair was reported and see why you’re wrong
And yeah, if it was disclosed it wouldn’t have been a big issue. But it clearly wasn’t. Want to make up different fact patterns and then discuss those, or discuss the situation we’re actually talking about?
That's irrelevant to what you're talking about. The Nvidia situation is its own thing, separate from your argument. Your argument is not about affairs, your argument is about people getting into any relationship with people they work with, in particular someone who may be of higher standing within a company. My claim is that, in most cases that is not as big a deal as you make it out to be like it's some caricaturized soap opera. Real life has real people in it and most of them can make their own adult decisions. In the vast majority of scenarios that people armchair moralize about, the actual people involved do not see a victim or feel victimized in any way, because it's mutual and not manipulative.
If you are a CEO who is personally responsible and directly respected for how much you've improved your company and some kind of scandal unfolds publicly, of course stock prices will drop. It's a signal that you might be stepping down or leaving the company, which in the case that you are such a high performing CEO that you personally have a huge impact on the company's future, can absolutely damage the future performance and stock value of the company.
Stocks don't care about "direct report" relationships or affairs or any of that as much as you think. If the whole spy angle hadn't happened, this would probably barely be public at all. The affair aspect makes it worse, but if the affair had only been with a Chinese honeytrap (if it's proven that she even was), you could honestly be forgiven for getting seduced by a nation-state level succubus. That becomes less of an excuse when you have a second affair with someone who isn't a spy. It's just its own thing and this power nonsense is mostly people addicted to victim mentality larping their own social warfare fantasy.
It's certainly a false equivalence to suggest that someone rich and smart, probably with alternative options and something to gain, is in a similarly vulnerable space as someone for whom losing their job means they're on the street by the end of the month.
I disagree with your argument as stated, as offering promotions in exchange for sexual isn’t uncommon, and many times isn’t really a choice.
But I also think this perspective isn’t really the point. Regardless of the abusive nature, it’s still a tremendous conflict of interest for an affair of this nature to occur. The direct report would have a huge amount of leverage over Huang to keep the affair secret for one thing.
I don’t know the nature of these particular affairs, but regardless of who is or isn’t vulnerable, it’s still wildly inappropriate, and shouldn’t be tolerated
How do you think executives that do work at this level meet people outside work?
The only people who understand this kind of job and what it takes are people in the same sort of area. And you’re prolly only meeting them in your own workplace.
At least if you’re boning someone from work you don’t have to watch everything you say the same way.
That's perfectly feasible and does not change the power dynamics at all.
I've seen people attracted to higher-ups ("coming on to them") in University and employment situations, and from outside looking in, the attraction could have been about the "power" the higher up was wielding (not necessarily using for bad; just the stature they had in the local society). TBH, people get attracted to people in power or with fame all the time (without actually knowing the person they are), so there is no reason it would be any different when there is a direct professional relationship.
Yes, this means it's tricky, and puts large demands on the "higher ups" — while they are similarly human with all their faults, it's still not necessarily the right thing to do — even, as a higher up, I'd struggle to differentiate if they are attracted to my "social standing" or to myself (if that can even be decoupled)?
exactly. Jensen had no option but to fuck this person. How could he possibly say no. What a unique angle that you came up with. Lets bring jensen back now.
HN being ugly is intentional. It actually does not need these facelifts, because its plain appearance serves as an important filtering function for who participates.
Yes and no. The US presidency is very different than the way it's reported on, but it is true that the executive branch is being more assertive to make up for the disfunction of the legislation. We're doing that intentionally and for good reason, to demonstrate to China that the US can act with speed as one small part of the overall deterrent against World War 3. The US is almost always hugely misunderstood, because there isn't any other country in the world that actually works exactly like the US does.
No one knows how the US works anymore because the current administration doesn't follow laws and just makes it all up as they go along. It turns out the court system is too slow to matter: illegal tariffs just move to new illegal justifications when the patently absurd ones are tossed out after more than a year of using tariffs to drive inflation and stroke Trump's ego.
How about ask AI for all the times all the US presidents did something that went against the constitution. I think you'll be shocked. The court system being a bit slow is part of the point, because we're in an environment where China wants to use the speed of democracy as a weakness. Since you're on HN maybe you understand the concept of slow journalism or the benefits of anything being slow. Most people agree with that, having things be a bit slower is a useful check against errors. We understand that, the US was literally founded on that. China does not care. In that environment, we have to make compromises.
Being less predictable makes it harder for our adversaries to calculate. We're still good people, but we're in bad times. You don't need to understand. Through most of history only a tiny number of people anywhere in the world truly understood what was going on and what was necessary to respond to the times. This is true in science as well. How many times in history was it only a few scientists that had a clearer picture? Everyone thinks their religion is _the_ religion. The same can apply to news and opinions.
If you think the US is about Trump's ego, you still don't understand.
I did ask and what I found out is that this admin is having nationwide injunctions at about 5x the rate of the previous admin, with 50 through July 2026. Biden had 28 in his entire term. Trump 2 is on pace to blow well past the previous record holder, which was the first Trump presidency at 80 total.
Moreover, Trump has made public comments regarding treaty obligations and international law that makes it clear he has no respect for those things, and there are no principles he follows at all so no one can predict what he will do next.
The principle is simple. Stop World War 3 by rolling up the vulnerabilities and dependencies that would give Chinese analysts a favorable picture. Forcing Europe to invest more in its own defense reduces the vulnerability there. Controlling oil going from Venezuela to China reduces China's sanction-proofing. Destroying Iran's space program, AI research, nuclear research, navy, airforce and most of its industrial base weakens their ability to cooperate with Russia and China as well as their capacity to contribute to global instability if they synced up with China. Helping Ukraine target Russian oil infrastructure also puts oil trade from Russia to China at risk, so they can see their own vulnerabilities.
Renegotiating the deal around Greenland prevents the risk of Greenland becoming independent and falling out of NATO since it would not remain in NATO by default.
It's not normal to capture heads of state, but Maduro was not the head of state as he was not recognized as elected by the world. It's not normal to kill heads of state, like Saddam Hussein or the leadership of Iran, but Iraq and Iran have something in common. Both Iraq and Iran attempted to assassinate US presidents multiple times, not to mention the mass killing of their own people. When Iraq attempted to kill a past US president, Bill Clinton had Iraq's special services HQ obliterated and no love was lost eventually when Saddam was hanged by his own people.
Learn more about what China has been doing in their region, around the world and inside the US. China thinks they can break all of the rules and then feign cute innocence like everyone else is crazy as if they are the only one following international law, when it is nonsense. We're operating right at the edge of the rules for sure. You won't find me disagreeing about that.
Our current foreign policy is the most sane it has been in a while, yet it looks a bit more extreme as a result of being so far behind on it. This is mostly what catching up looks like, when being polite doesn't get the job done fast enough.
That's a nice cherry picking to paint a narrative, but the US is trying to force Ukraine to stop attacking Russian oil infrastructure and putting tariffs on allies instead of just China and chinese allies pushes the world towards them.
The Greenland deal was always going to be renegotiated as it always has been, once again antagonizing allies was completely unnecessary.
The US is not operating at the edge of the rules, it is constantly breaking them and tearing down any respect that the world still had for the rules.
At this point I don't know why the world should even have a more positive view of the US than of China. China is far more peaceful and helpful.
The US doesn't have a problem with Ukraine hitting Russian oil infrastructure in general, just specific oil infrastructure that happens to hit US interests. As I said, we've been helping them do it. It's partially political and around diesel fuel specifically, which is not all of the oil Russia produces.
None of these countries want to be more reliant on China. While China may be benefiting from their propaganda against the general populations, western leadership is a little more clear eyed about the risks of China. It's more for China to lose if they try to take Taiwan.
Whether the rhetoric around Greenland was entirely unnecessary or not, it may have accelerated the timeline on getting a deal done and this is Trump's negotiation style since he's been doing it forever. It's also interesting strategically, because if you'll recall China's brand of communism was only partially class warfare, but was also anti-colonialism. Framing all such foreign interests as colonial acts even though that is not what we're doing, paints any action by China that looks similar to Cube or Venezuela (which are 2 variants of scenarios we have theorized China might use against Taiwan) will inevitably draw comparisons, even though the details around our actions are very different than the overall rhetoric we're using.
The rules aren't gone. If you look at the rules the US has historically tried to enforce in terms of international law, we're largely still doing that. Do people think that Russia, China, Iran and Venezuela have been following and promoting international law? No. They've been actively and aggressively breaking them, because they see it as a strategic advantage to do so. If we make it impossible for us to respond to their rule breaking, by not getting creative with the rules ourselves, then we allow them to legitimize their rule breaking as the new international norm. Putin says hey, don't do that, hey you can't do that! China says hey, that's against international law, we're stable, you can't do that! It's only to serve their own interests. Of course there are red lines, but they can't say we're crossing red lines of behavior that they themselves are engaging in. That is the game.
If you legitimately don't understand why China is bad, then learn. You're running out of time to do so.
Chinese analysts consider the network of alliances that the US has built since WWII a key advantage over China. Nothing could be more favorable to China than Trump openly attacking the very concept of NATO as a platform for mutual cooperation and questioning US commitments to South Korea and Japan.
After years of mocking the sovereignty of allies like Canada and Denmark, are they more or less likely to support the US in the event of a war with China? Less likely, and the same applies to most other countries, too.
Exactly, and our allies aren't going away. There is no world arriving where these countries have more in common with China than the US. NATO isn't going anywhere either, that's all rhetoric. Go subscribe to NATO news and see how they talk about things. Go watch the many hours of recordings of what we say directly to NATO with our representatives. You'll come away with the exact opposite impression that you would get from popular opinion news. They are investing more into their own defense and industry. If China alienates them, then they will be a little more prepared to contribute to a World War 3 if it comes to that.
Chinese analysts have a long history of being hilariously wrong. After all, communism, merit based. Pick one.
Maybe you'll notice that Canada has been a bit leftist for a while. Maybe you'll notice that historically if a country's sovereignty is questioned or nationalist sentiments increase, these countries begin to lean back to the right. Maybe you'll notice we've been doing that intentionally around the world to make the world more conservative. Globalism went too far and for strategic reasons it would be beneficial to unwind that some. These are basic things. It has the extra benefit that China has less influence with conservatives, so by increasing the influence of conservatives around the world, we are simultaneously reducing China's influence in those places.
Maybe you'll notice that we haven't actually invaded Greenland or Canada. Maybe you'll notice that the countries we're actually physically acting against are dictatorships, not democracies.
We don't need you to understand. You will feel all your emotions and your emotions will control you and you will behave exactly like we're banking on you to. The reality is that the US remains the only superpower despite China's advancements and we're still using that power for good, whether you realize it or not.
NATO allies are already pivoting away from the US, and the French nuclear deterrence initiative, which is gaining steam, is a good example of that: https://www.wsj.com/world/europe/frances-nuclear-forces-brin... The current military stack that heavily depends on the US is being slowly and methodically replaced with purely European solutions. This means that Washington will have less and less say in how and when it is used.
Anything US-related is increasingly seen as a liability. Take the HIMARS, for example. An increasing number of countries that recently bought HIMARS are favouring South Korean Chunmoo for their next purchases. Chunmoo comes with technology transfer that allows customers to maintain the launchers and manufacture the ammunition themselves, whereas HIMARS is micromanaged from Washington, and poorly so.
> Maybe you'll notice that we haven't actually invaded Greenland or Canada.
The threat to Greenland got real enough that Denmark moved special forces to Greenland, rigged airport runways with explosives to disrupt US airborne assaults, and stocked up on blood to treat casualties. Only then did the rhetoric from Washington change. That burned through most of the goodwill Denmark (and many of its supporters) had towards the US.
> The reality is that the US remains the only superpower
Doubtful if the US deserves to be called a superpower anymore. Instead of supporting Ukraine to the fullest extent possible and setting an example that America's partners are not to be messed with, the US government has consistently shown weakness by falling for Putin's bluffs and humiliating itself in the process. Witkoff calling Putin a "humanist" is a good recent example. Pathetic beyond words.
The US is losing credibility as a security guarantor, because it has failed to impose its will even on countries such as Iran. From failing to adopt modern anti-drone warfare, and using very expensive interceptor missiles instead, and then running out of them, we can all see that the US military and industrial might is a postcard from the past.
Why does China even need to fight or even directly confront the US when it can simply bribe its government into halting military aid to Taiwan, pressuring Taiwan to surrender, and pressuring others to drop their sanctions? Putin offered Lukoil to Trump's buddies, and today's WSJ top story says that Trump is pressuring Europe to resume buying Russian gas so that his friends can use intermediaries like Lukoil to skim their share of the revenue and enrich themselves. China can offer TSMC to Trump's buddies, too. That would send decades of US wargaming down the toilet, just as Trump's actions have done regarding Iran and Russia.
> The current military stack that heavily depends on the US is being slowly and methodically replaced with purely European solutions.
Which is exactly what needs to happen.
> An increasing number of countries that recently bought HIMARS are favouring South Korean Chunmoo for their next purchases.
I don't see a problem with this.
> The threat to Greenland got real enough that Denmark moved special forces to Greenland, rigged airport runways with explosives to disrupt US airborne assaults, and stocked up on blood to treat casualties. Only then did the rhetoric from Washington change. That burned through most of the goodwill Denmark (and many of its supporters) had towards the US.
That sounds like performative politics. It says nothing about how real the threat actually was and is more likely a performative political demonstration to Greenland that being tied to Denmark has value, which would be a net win for the US since a large part of our concern about Greenland was that they were being given the option to become independent. So performative politics for Greenlanders that has strategic benefit to the US more than Denmark, or useful military practice. It doesn't make it any more real or not.
> Doubtful if the US deserves to be called a superpower anymore. Instead of supporting Ukraine to the fullest extent possible and setting an example that America's partners are not to be messed with, the US government has consistently shown weakness by falling for Putin's bluffs and humiliating itself in the process. Witkoff calling Putin a "humanist" is a good recent example. Pathetic beyond words.
We don't need the US industrial base ramping up for the Ukraine war. We need it ramping up for the war with China. Europe needs to be ramping up for the European war.
As far as Steve Witkoff goes, he's literally a special diplomatic envoy. The point of a diplomat is not to go and insult the people you're communicating with, it's to achieve results through diplomatic means. Saying positive things is also a way to get your message into Russia past the censors.
If you let the media tell you what to think and feel, then they will preserve your lack of understanding of what is happening which creates a self reinforcing cycle of you depending on their nonsense.
> Which is exactly what needs to happen. /---/ I don't see a problem with this.
Representatives of the US government do not share this sentiment and throw temper tantrums all the time. One example of many:
BRUSSELS — U.S. Deputy Secretary of State Christopher Landau on Wednesday slammed European NATO allies for prioritizing their own defense industry over American arms suppliers, according to three NATO diplomats. /---/ Landau, a longtime NATO skeptic who spoke first at the closed-door meeting, told ministers not to “bully” his country’s defense firms out of participating in Europe’s rearmament.
The expectations seems to have been that elevated military spending would flow to US arms manufacturers. Instead, it is flowing to European manufacturers, making them more competitive and reducing the influence that the US government had through export controls, which essentially gave the US veto power over the foreign and defence policies of other countries. That influence is now rapidly waning.
> That sounds like performative politics.
Certainly not. This was an European country preparing for an attack by a NATO ally. The forces on Greenland were issued live ammo and explosives to sabotage airports to prevent US airborne attacks from establishing bridgeheads, and ordered to engage the invading force without further instructions from the high command. Nothing about this is performative, especially given that it was not revealed to the public at the time.
This was the moment when trust was broken beyond repair. No amount of flowers or apologies will restore the trust that was before. It is perhaps one of the most important events of the past decade in US-EU relations and a fundamental shift in the understanding that had stood since 1949 when NATO was established.
> We don't need the US industrial base ramping up for the Ukraine war. We need it ramping up for the war with China.
And how's it going? 400 Patriot missiles were manufactured in 2022. After years of "ramping up production", 620 were manufactured in 2025. The US inventory at the start of 2026 was 2300 missiles, 1700 of which were pointlessly wasted on Iran. It will take years to restore even the fairly small inventory (judging by how many were consumed in the war) that was at the start of the year, let alone anything needed for a major war. The picture with dumb munitions like 155mm artillery shells isn't much better.
> The point of a diplomat is not to go and insult the people you're communicating with, it's to achieve results through diplomatic means.
A fair bit of latitude separates insulting someone from publicly licking their asshole and going, "Yum! So tasty!"
There's a reason why the US government has not achieved any meaningful concessions from Russia since the 2021 ultimatum. The endless talks with Witkoff serve only as a humiliation ritual. Putin even got to walk down a red carpet on US soil while US officers saluted him, all without giving anything in return. Allegedly, that embarrassing Alaska summit happened because Witkoff misunderstood what Russians meant by "withdrawal from Zaporizhzhia and Kherson". He mistook Russian demand for Ukraine to withdraw from Zaporizhzhia and Kherson for an offer by Russia to withdraw its own forces. This led Trump to immediately annouce the summit in Alaska that produced flashy photos for Russian newspapers, but nothing else. My God, what a bunch of retards. They don't even understand what's being told to them, and you're trying to condescendingly convince me that they're playing clever reverse psychology on Europe. It's time to face the reality.
Landau can have his opinion, and I think he mostly wants US industry to have a fair chance, but we do need Europe to build it's own capacity.
Purchases from US manufacturers have increased, which contradicts your claim that they have not. Europe industrial capacity is also increasing. So, this all sounds fine to me.
I think you vastly overestimate the importance of the Denmark and Greenland thing in terms of meaningful harm to US interests. This is very basic bread and butter foreign policy. Just because it's not what you're used to and makes you uncomfortable doesn't mean it is actually wrong. Your feelings are worthless and reality gets all of the votes.
Re patriot missiles, there are a few things you need to understand. They are not the only kind of defense system we have. We do not publish actual numbers, they are secret. The way our defense acquisitions work is kind of political, so we have to get Congress to invest into any particular needs.
We don't need to scale up patriot missiles anyway, because they aren't the defense system of the future anymore. They have specific uses, but they've been getting used for purposes outside of their optimal. They're too expensive and complicated for the way the world is going.
So everyone has tried things that don't work with Russia. Somehow Trump failing at it in his own way is meaningful? Everything is emotion emotion emotion with you. It's about how you feel, how you project human emotion onto geopolitics. You're wasting your time trying to do that. Oh we should be so ashamed! The drama, the scandal! How inappropriate! The end is near!
Move beyond tabloid thinking and the world will make a lot more sense. Like I said though, we don't need you to understand and most people never will. It's fine.
Did y'all do any particular work that makes these models great for moderation? One of the issues between cloud models vs local is that often people are trying to push the limit of what their hardware can do and it doesn't leave too much headroom for moderation to make the models a little more constrained for kids or schools to use locally, or just for family in general who don't want accidental unhinged results.
In the past I've noticed that some LFM were way better at moderation than much larger models. Not necessarily prejudging what is acceptable, but allowing you to interrogate content with a series of questions to customize the moderation. Other models I tested were far worse.
Decision models are like the next optimization in that direction, but none of the benchmarks for the popular models ever show how good they are at tasks like content moderation. The smaller and more efficient we can get that workload, the more that it can simply be on by default in a lot of use cases without feeling like you're giving up a lot of performance, creative freedom or accuracy. You can't have a high number of false positives or people would rather take control back.
We need local content moderation to be cool, efficient, accurate and configurable so that people can get the higher quality results which unsafe models tend to produce, while filtering out the unsafe results fast enough that it can even be done in an agentic way and fed back into the model to steer its results.
on moderation, we run an open guardrail benchmark at raxIT Labs and the hosted d1 models do well. d1.1 is #1 of 30 at checking model output for harmful content (94.4 balanced accuracy), d1-ultra is #1 on harmful user requests (88.2). we haven't run the new 2B open weights yet so no number for those. results and method: https://dm-as-guardrails.raxitlabs.com
This tracks with my experience with their models. Nice work. It would be interesting to see what they could do if they optimized even more for moderation like this without going so far as to hurt the model performance overall.
Can you describe more in terms of what moderation you want? I can say with confidence that we don't train for the typical adult-only content you can see on some "de-safe'd" models, but it's not clear to me exactly what you're asking about. We can also discuss over email if you prefer.
I'm largely referring to the fundamental components that can be combined to arrive at a decision. Is this a real life image or text, or is it some sort of art or fiction? Is this in a context of historical events for education, etc? Is this sarcasm? Is this a real threat? Is this nudity? Does this contain violence and gore? Is this suggestive? How old is this character? Is this content sufficiently related to the content it's replying to, or is it spam? Does this response deviate too much from the input request? Is this a business logo?
There was this book called The Inquisitive Mood and it very much reminds me of the chain interrogation process you can do with models now.
This isn't exactly a single use case, it's more like there are use cases for image editing software that use generative models, there are use cases for local LLMs to get a sense for abstract user intent, or to check if a model's action is in line with policy and so on.
As I said, the Liquid AI models are already very good at a lot of this, but considering how good the models are at it, I'm curious how much y'all train on tasks that are moderation-adjacent. If y'all don't have any model that's really specifically trained for these kinds of tasks, it would be interesting to see how close y'all could get if you tried.
This is an actually important problem in the AI space especially for local development.
There are other moderation tasks that are harder, like "is this copyrighted?" or "is this a McDonald's logo?" because of how much copyrighted material there is and how many businesses there are, but it might be possible to get a good feel for "is this likely a trademark?", "is this a business logo?", "does this text seem to describe a popular character?"
To some extent you can combine a model like this with a good database to optimize the filters and identification. Then you can provide a few reference image or text examples that are closer to what you're trying to filter out and the model can say, "yeah, this content is pretty close to your examples."
Google and OpenAI have already released some of their open source safety models and they were really quite bad, from my testing. Yet, Liquid AI hasn't really leaned into promoting their models for moderation purposes despite being excellent. I know there is a lot of political intersection in moderation, especially in the past many years, but I think a model that is smart enough and excels at letting people apply their own rules with fairly good accuracy is important right now.
This would be amazing for making audiobooks more expressive and like real storytelling. I kept telling Audible they needed to invest more into premium audio experiences, kind of the way you had the old radio plays that people used to tune into. The War of the Worlds! We are seeing some of that with full casts of actors doing readings more.
At the same time, imagine if every book was like a living thing you could interact with and buying the book partially bought you access to the model that would tell you the story. The only way you would interact with that particular model, is if you bought the book.
It reads the book to you and you could chime in or it would know when to pause if there's an interruption without you needing to manually pause. You say hey wait, why is this character so angry? Then it kind of naturally reacts and continues the story. Maybe you come back to it 3 years later and it remembers you liked a certain part of the book and injects "here comes the good part!" It hears you start snoring and begins lowering its voice and whispering, but remembers you asked it to wake you up loudly if you start snoring, so it bolts out HEYWAKEUP!
For a lot of people books are a very solitary thing, though of course there are social clubs and book readings. Almost makes me think of those magical books you see fantasy wizards having, where the books themselves are sentient. It would be crazy if they released a The Princess Bride book and had the real narrator's voice and you could interrupt him and he could harass you about all your interruptions.
It wasn't that long ago where we were talking about having games and movies blend so the movies become interactive experiences. We're really getting closer to some of these static mediums becoming more like, trained behavioral meshes that have a personality and you can kind of naturally spend time with it. I'm not sure I'd be big on paintings on the wall coming to life and talking to you like in Harry Potter, but there's some promise.
Game worlds are an obvious extension, not just through NPCs, but if anyone remembers that game called Bastion where there was a narrator explaining the events as they went on. You could imagine a model that could watch you play and speak naturally, being funny or narrating. Almost something like Mystery Science Theater 3000 or how Beavis and Butthead would react to music videos.
I don't think those should be pervasive or the new norm, but there probably is an opportunity there and a market for it. Some kind of social-like presence that doesn't get tired, doesn't get bored, always has the same schedule you do, etc.
Hell, that reminds me of Seaman on the Dreamcast, too. SEGA could make a new Seaman and it would actually be awesome. Like a blend between Nintendogs, a fish, and some AI. There could even be charming goofy antics it tries to pull where it knows it's stuck inside a game and tries ridiculous things to manipulate you into letting it out. Some kind of Pinky and the Brain charm.
This is not necessarily the case, as you still need to deliver an experience that customers enjoy and will pay money for. This means consoles that let them play games in their hands with the highest picture quality and lowest latency. No disruptions from poor wireless or significant interference in environments like apartment complexes.
Fully homomorphic encrypted processing may also become a norm and if that is combined with some games going back to being console only, there could be significant enough protection for games that contain trade secrets. That said, the days may be numbered for trade secrets at all given that better models may make today's clever trade secrets look juvenile.
At the same time, a lot of these games were already getting decompiled and inspected at runtime. That has historically been one part of the incentive to release games on console only or console first, particularly first party titles.
In any market where it's relevant to their business interests, they can still rely on the copyright laws and can set traps for any companies using decompiled code as part of their own paid and sold product. That whole process will become cheaper and easier to do as well, plus new approaches to slowing down or hindering the results of decompilation will become normal for games in mathematically provable ways. We've already seen approaches over the years like Denuvo.
Always online games themselves have benefited from some of their code being in the cloud and that will obviously be under fire more than ever with people writing their own servers, but even that might have novel approaches that can survive advanced tampering.
Indie platformers will continue to run entirely on the console. But AAA games have a history of prioritizing visuals over everything else, because that sells trailers. Fully streamed games executing on a remote GPU allow better graphics and prevent decompilation and piracy, all great things from the publisher's POV. So far the economics haven't been appealing and the experience isn't uniform for customers, but I could see AI-assisted decompilation pushing the AAA publishers over the threshold
If you can't profit off of your product sufficiently, because not enough people are interested in the experience that streaming provides, then you can no longer justify even making those games and it all becomes irrelevant. So, in the end you release your game to local hardware anyway.
They can't simply copy your game and reship it, because that would be against the copyright laws in most places where these companies make their profits from. Making something sufficiently new would be technically legal, but it would also take massive computing resources even with LLMs. That's either local compute time or cloud spend.
I think the bigger risk is just brand new games being generated of all kinds rather than the more expensive and legally risky full decompilations. Mass generation could start saturating the market and soaking attention. The low end indie market is already pretty saturated and in the end games have to be appealing, review sufficiently well and find an audience. Plenty of good games have barely sold any copies.
So almost a bigger question will be, how will AI impact the marketing industry?
I've tried it some, only locally. I rarely use APIs, so can't say what the official inference API results look like. It correctly followed some prompts that almost every model I've ever tried performed poorly at, but failed in strange ways on prompts that other models handle with ease.
At high resolutions, structural integrity on people can definitely fall apart like older models, where you get one person turning into a long spaghetti spanning across the frame. It might be that you need to do grounding by generating some kind of scene reference at its stable resolutions first then having it use that for the larger resolution like older models too.
They say that their prompt "upscaler" model or system prompt are essentially necessary for the model, but my results have been hit or miss with that claim. Some basic prompts are better, some upscaled prompts are better. I think that if you gain an intuition for how the model understands things, you might get better results by default by hand crafting prompts specifically for it.
Some of the stylized art I'm getting out of it at higher resolutions are really nice and I prefer them over other models.
All of these models seem to have their own character in terms of system requirements, performance, prompt adherence, artistic outputs and so on. None of the models I have tried are a complete homerun, so maybe in this case the model feels a little inconsistent in some ways, but it's a model I will probably keep around.
Is it that they don't want to overstate their claims to investors who have to decide whether to actually build these, in case the results they're predicting don't add up?
Either way, we need to be working on EDAI, so if they've got something, hopefully it's something.
reply