On Sunday, OpenAI CEO Sam Altman offered two eye-catching predictions about the near-future of artificial intelligence. In a post titled "Reflections" on his personal blog, Altman wrote, "We are now confident we know how to build AGI as we have traditionally understood it." He added, "We believe that, in 2025, we may see the first AI agents 'join the workforce' and materially change the output of companies."

Both statements are notable coming from Altman, who has served as the leader of OpenAI during the rise of mainstream generative AI products such as ChatGPT. AI agents are the latest marketing trend in AI, allowing AI models to take action on a user's behalf. However, critics of the company and Altman immediately took aim at the statements on social media.

"We are now confident that we can spin bullshit at unprecedented levels, and get away with it," wrote frequent OpenAI critic Gary Marcus in response to Altman's post. "So we now aspire to aim beyond that, to hype in purest sense of that word. We love our products, but we are here for the glorious next rounds of funding. With infinite funding, we can control the universe."

AGI, short for "artificial general intelligence," is a nebulous term that OpenAI typically defines as "highly autonomous systems that outperform humans at most economically valuable work." Elsewhere in the field, AGI typically means an adaptable AI model that can generalize (apply existing knowledge to novel situations) beyond specific examples found in its training data, similar to how some humans can do almost any kind of work after having been shown few examples of how to do a task.

According to a longstanding investment rule at OpenAI, the rights over developed AGI technology are excluded from its IP investment contracts with companies such as Microsoft. In a recently revealed financial agreement between the two companies, the firms clarified that "AGI" will have been achieved at OpenAI when one of its AI models generates at least $100 billion in profits.

Tech companies don't say this out loud very often, but AGI would be useful for them because it could replace many human employees with software, automating information jobs and reducing labor costs while also boosting productivity. The potential societal downsides of this could be considerable, and those implications extend far beyond the scope of this article. But the potential economic shock of inventing artificial knowledge workers has not escaped Altman, who has forecast the need for universal basic income as a potential antidote for what he sees coming.

Criticism of predictions of impending AGI

Artificial workers or not, some people have already been calling "BS" on Altman's optimism. It's nothing new. Marcus, a professor emeritus of psychology and neural science at New York University, often serves as a public foil to Altman's pronouncements, a trend that largely began when Marcus appeared before the US Senate in a May 2023 hearing as a skeptical counterpoint to Altman's testimony during the same session.

On Sunday, Marcus laid out his most recent criticisms of OpenAI's prediction of achieving AGI soon in a series of posts where he detailed how current language models sometimes fail at basic tasks like math problems, "commonsense reasoning," and maintaining accuracy when faced with novel problems.

OpenAI's current "best" released AI model, o1-pro, what you might call a "simulated reasoning" or SR model, reportedly performs well on some mathematical and scientific tasks but still shares weaknesses with OpenAI's GPT-4o large language model, such as failing to generalize well beyond its training data. And it may not be as strong as OpenAI claims in some cases.

For example, Marcus cited a recent benchmark conducted by All Hands AI that reportedly shows that OpenAI's o1 model scored only 30 percent on SWE-Bench verified problems (a set of GitHub-based problems), which is below OpenAI's claimed 48.9 percent performance rate, while Anthropic's Claude Sonnet (which is not purported to be an SR model) achieved 53 percent on the same benchmark.

Even so, OpenAI claims further progress on its AI model capabilities over time. In December, OpenAI announced o3, its latest SR model that impressed some AI experts by reportedly performing well on very difficult math benchmarks, but it has not yet been released for public examination.

Superintelligence as well?

Altman's post follows his September prediction that the AI industry may develop superintelligence "in a few thousand days." Superintelligence is an industry term for a hypothetical AI model that could far surpass human intelligence. Former OpenAI Chief Scientist Ilya Sutskever founded a company around the pursuit of the technology last year.

Altman addressed the topic in his latest post as well.

"We are beginning to turn our aim beyond [AGI], to superintelligence in the true sense of the word," he wrote. "We love our current products, but we are here for the glorious future. With superintelligence, we can do anything else. Superintelligent tools could massively accelerate scientific discovery and innovation well beyond what we are capable of doing on our own and in turn massively increase abundance and prosperity."

Despite frequent and necessary skepticism from critics, Altman has been responsible for at least one verifiable tech catalyst: the release of ChatGPT, which served as an unexpected tipping point, he says, that brought AI to the masses and launched our current AI-obsessed tech era. Even if OpenAI doesn't get to AGI as soon as Altman thinks, there's no doubt that OpenAI has taken the technology to unexpected places and spurred wide-ranging research on AI models in the tech industry.

"We started OpenAI almost nine years ago because we believed that AGI was possible and that it could be the most impactful technology in human history," he reflected in his post. "At the time, very few people cared, and if they did, it was mostly because they thought we had no chance of success."

Read full article

Comments

It's getting harder to access popular adult sites in the US South.

On Wednesday, Pornhub's owner, Aylo, kicked off the new year by blocking two more states that implemented age verification laws requiring ID to access porn, Florida and South Carolina. According to 404 Media, these states are now among 16 states where Aylo sites, including Pornhub, RedTube, and YouPorn, cannot be accessed. Tennessee also risked being blocked, but a court preliminarily blocked its age-verification law from taking effect.

The other blocked states are Alabama, Arkansas, Idaho, Indiana, Kansas, Kentucky, Mississippi, Montana, Nebraska, North Carolina, Oklahoma, Texas, Utah, and Virginia. Mapping it out, 404 Media noted that the Aylo blackout spans nearly the entire US South, with Georgia's age verification law set to take effect in July and likely to trigger another block that would almost complete the blackout.

When users in any of the blocked states attempt to access sites like Pornhub, they see a message from adult performer Cherie Deville informing them that access is blocked for their own safety.

Aylo has long argued that age verification laws intended to protect kids from accessing adult content are flawed. Rather than keeping kids away from adult materials, these laws actually serve to restrict adult activity online, Aylo argues.

Aylo also says that requiring ID simply drives adult users to riskier adult sites that won't comply with laws or moderate content wrongfully uploaded without performers' consent, as Pornhub and other Aylo sites claim to do. And requiring adult sites to store personal information about each user also poses privacy risks, Aylo warned in a statement to 404 Media.

"First, to be clear, Aylo has publicly supported age verification of users for years, but we believe that any law to this effect must preserve user safety and privacy and must effectively protect children from accessing content intended for adults," Aylo said. "Unfortunately, the way many jurisdictions worldwide, including Florida, have chosen to implement age verification is ineffective, haphazard, and dangerous. Any regulations that require hundreds of thousands of adult sites to collect significant amounts of highly sensitive personal information is putting user safety in jeopardy. Moreover, as experience has demonstrated, unless properly enforced, users will simply access non-compliant sites or find other methods of evading these laws."

Perhaps the easiest way to evade state age verification laws is by using a VPN to mask a user's location. According to a report from VPNMentor's cybersecurity and research lab, VPN use in Florida jumped by as much as 1,150 percent within hours of the age verification law coming into force on January 1. That's even higher than the 967 percent spike in VPN use that VPNMentor saw in Utah in 2023 and substantially higher than a Texas spike of 234 percent—perhaps indicating that users are getting savvier at quickly securing VPN workarounds.

Some First Amendment advocates, including the Free Speech Coalition, have joined Aylo's fight and are pushing the Supreme Court to block Texas' law. Many digital rights advocates and First Amendment scholars have filed briefs in support of the Free Speech Coalition, which has urged the Supreme Court to agree that Texas' law is not narrowly tailored and "burdens vast quantities of speech protected for everyone." If advocates win, the Supreme Court may order an injunction that could impact all states that passed essentially copycat legislation, potentially ending Aylo's long war against "haphazard" age-verification laws in the US.

In its most recent filing, Texas argued that only intermediate scrutiny should be applied, pushing SCOTUS to agree that Texas has no other remedy to protect kids online. Texas claims that if Pornhub can verify the ages of content providers, it should have no problem verifying the ages of viewers.

Oral arguments in that case begin on January 15.

This post was updated on January 2 to clarify that Tennessee has not been blocked.

Read full article

Comments

Nintendo Switch 2, which may not even be called that, will be released in 2025. We think. It will launch with some games, but we don’t know which ones. We’ve got a relatively complete, if hazy, picture of what the console will be — a new, more powerful Switch, basically — but we don’t know what, if anything, will distinguish it from its predecessor other than that.

There’s no doubt that the Switch 2’s launch will be one of two seismic events in gaming in 2025, alongside Grand Theft Auto 6 (assuming Rockstar’s game makes its fall 2025 release window). But for a console that could reportedly launch as soon as March or April, it’s remarkable how much about the Switch 2 remains uncertain: the details, certainly, but also the bigger picture. What’s at stake for Nintendo? What could go wrong? And how will the rest of the industry react?

Let’s start by recapping what we know so far, and what we don’t, before asking some of those existential questions.

Here’s what we definitely know about Switch 2

This is everything Nintendo has said on the record about its next console:

• It exists
• It will be backward-compatible with Nintendo Switch games
• It will offer the Nintendo Switch Online service
• It will use Nintendo Accounts
• Nintendo will officially announce the console within its current fiscal year, so before the end of March 2025
• That’s it!

Here’s what we think we know about Switch 2

Through well-sourced reporting and leaks, a fairly clear picture has built up of the new machine in advance of its unveiling:

• It’s called Nintendo Switch 2.
• It will be released in 2025. Nintendo had planned to release it before the end of March, but there are indications this deadline has slipped, with Nintendo keen to avoid stock shortages and to have launch software ready.
• It will be a handheld hybrid console with detachable controllers and a dock for connecting it to a TV, just like the original Switch. (This is strongly supported by the backward-compatibility announcement.)
• It will be compatible with some Switch accessories, including original Joy-Con controllers (even if it won’t be able to connect to them physically or charge them).
• The Switch 2’s Joy-Cons will attach magnetically, rather than using the Switch’s slide clip system, and feature Hall-effect sticks and some extra buttons.
• It has a cartridge slot like the Switch’s, and will support physical releases.
• It’s larger, with an 8-inch 1080p LCD screen.
• In terms of power, it will be close to the Xbox One/PlayStation 4 home console generation. But with custom Nvidia hardware that allows for machine-learning-powered upscaling and even ray tracing, it could get surprisingly close to the Xbox Series X and PlayStation 5 experience.
• The base onboard storage will be 256 GB, expandable via microSD card. It has two USB-C ports.

Here are some pretty solid guesses about Switch 2 games

Notably, there have been no leaks or well-sourced reports about the Switch 2’s software lineup. But knowing what we do about internal development at Nintendo, we can make a few informed guesses.

• Metroid Prime 4: Beyond, confirmed as a Switch game, might actually be a cross-generational launch title for the Switch 2, in the manner of The Legend of Zelda: Breath of the Wild (Wii U and Switch) and The Legend of Zelda: Twilight Princess (GameCube and Wii).
• Some first-party Nintendo games that are not necessarily launch titles, but probably arriving early in the life of the machine: a new Mario Kart, a new 3D Mario title, a new Animal Crossing game, and long-rumored Zelda remasters The Wind Waker and Twilight Princess. We can make these guesses because of the conspicuously long gaps to the last releases in some of these franchises; it’s been a decade since there was an all-new Mario Kart, seven years since Super Mario Odyssey, and four years since the 46-million-selling Animal Crossing: New Horizons.

Here’s what we definitely don’t know about Switch 2

• How much it will cost
• The exact release date
• Any confirmed games (beyond a couple of minor third-party releases)
• What, if any, gimmicks it will have to differentiate it from the Switch

Seriously though. How much will it cost?

This is the big one. Nintendo is in a good position; the Switch has built up a massive install base for the company, plus Nintendo has all its beloved properties and decades of trust in its brand to lean on. But, like all tech companies right now, it’s under intense pressure on pricing, and the wrong move could spell disaster.

The Switch launched at $299 in 2017, and its official sticker price hasn’t dropped a single cent since. This is in line with the changing trends in pricing — due to chip shortages, inflation, and the slowing pace of advances in tech, this generation of consoles has not come down in price as much as previous generations did. (It’s also true that demand for Switch has remained so consistent that Nintendo hasn’t needed to drop prices, and it has the Switch Lite to market at a lower price point.)

In this environment, Nintendo will have to figure out how to package more advanced tech while keeping the price reasonable. The Switch 2 is unlikely to be priced any lower than $349, the current cost of an OLED Switch model. $399 seems like a safe bet — the same price as the base Steam Deck. Any more than this and Nintendo will face uncomfortable comparisons to the new wave of PC handhelds (a Steam Deck OLED is $549; an Asus ROG Ally is $499) and risk pricing itself out of its traditional family market.

What’s at stake for Nintendo?

Almost everything. As I said, Nintendo is sitting pretty right now, but the games market is changing rapidly. The cost of producing games is soaring without market growth to pay for it, margins are being eaten into everywhere, and traditional console hardware platforms are losing influence. Nintendo’s business model — selling hardware and exclusive games to play on it — is emblematic of the old ways, but even Sony publishes on PC now. If Switch 2 goes wrong, Nintendo will probably have to face the ultimate indignity of becoming a third-party publisher.

At the same time, Nintendo is better protected against the winds of change than anyone else. It has vast cash reserves, incredibly valuable IP, growing multimedia clout, and a resourceful approach to technology that helps keep both hardware and software costs down. Perhaps it could weather another generation on its uppers.

Nintendo has also usually taken an innovative approach to its hardware platforms that has worked out more often (Wii, DS, Switch) than not (Wii U). But this is not necessarily the time for innovation. 

The Switch was an attempt by Nintendo to lock down its dominant position in handheld gaming while keeping one foot in the home, and it was an amazing success. Since then, Valve, with Steam Deck, has come at the same problem from the other direction — what if you could take your home game library on the go with you? — and proven that it’s possible and that there’s a high demand for it. Now Microsoft and Sony are both reportedly planning handheld consoles.

The dedicated handheld space is about to become more aggressively competitive than it has ever been, and Nintendo needs to use the Switch 2 to dig in and defend its position. That’s why it has already announced backward compatibility, so the owners of the 140 million Switches Nintendo has sold can rest easy that their libraries will persist. For perhaps the first time in the company’s lifespan, continuity is everything, and it can’t afford to wander off on a tangent.

That’s why the Switch 2 sounds as though it will be Nintendo’s most conservative design since the GameCube. It simply has to succeed.

In addition to the git command line tool, there are other tools or services that let you view changes in git history. The most interesting cases are those which present changes as part of a pull request, since those are changes you are reviewing and approving. But a common problem is that what they show you might not be what actually changed.

I’ll limit my discussion to services and tools I have experience with, which means that it’s the git command line, Azure DevOps, GitHub, and Visual Studio. You are welcome to share details for other services that you use, particularly those used for code reviews.

First, let’s consider a commit that changes the encoding of a file. For concreteness, let’s say that the file is this:

I just checked.
It costs A31.

where A3 represents a single byte with hex value 0xA3. This is the representation of £ in the Windows 1252 code page.

Suppose you change the encoding of this file to UTF-8:

It costs C2A31.

If you view this in the command line with git show you get

The command line version shows you that there used to be a byte 0xA3 but now there is a £ character.

Next up is GitHub. Its diff says

GitHub assumes that all files are in UTF-8, so it interprets the A3 as an illegal UTF-8 code unit sequence and represents it with U+FFFD REPLACEMENT CHARACTER.

Next up is Team Foundation Services Visual Studio Online Visual Studio Team Services Azure DevOps. Azure DevOps. That’s the name. Azure DevOps.

Here’s what Azure DevOps shows:

The file differs only in whitespace.

And if you expand the file and enable “Show whitespace changes”, it shows you no changes, not even whitespace changes!

This is quite concerning, because it means that if you made a change to the text of a file and also changed the encoding, Azure DevOps highlights the text changes, but does not give any indication that the encoding changed!

For example, maybe somebody changed the first line of text and accidentally changed the encoding from 1252 to UTF-8. Azure DevOps shows this as

It happily shows you the text change, but completely ignores the encoding change.

That encoding change might have caused you to inadvertently change a bunch of strings in a Resource Script, resulting in mojibake.

If you ask Visual Studio to view the diff, it indicates that the file has been modified (M), but when you ask to see the diff, it says “0 changes”, and nothing is highlighted.

Now let’s consider a commit that inserted a UTF-8 BOM at the start of a file.

From the command line with git, you get this:

The BOM displays as a space. Not great, but at least there is a +/− to show you that something changed, and if the first line is not otherwise blank, the shifted contents tell you that something got inserted at the start of the file.

For GitHub, the diff shows up like this:

The highlights tell you that something changed on that line, but squint all you want, you don’t see any change. The change must be invisible, but at least you’re told that there’s a change somewhere on that line; you just can’t see it.

And finally, we have Azure DevOps:

The file differs only in whitespace.

As before, even if you expand the file and enable “Show whitespace changes”, you get no changes.

So Azure DevOps tells you that the file changed in whitespace, but when you ask to see it, you are shown no changes.

If you ask Visual Studio to view the diff, it once again indicates that the file has been modified (M), but when you ask to see the diff, it says “0 changes”, and nothing is highlighted.

I suspect that in the cases where GitHub, Azure DevOps, or Visual Studio show no visible changes, most users will just conclude, “Must be a bug,” and not realize that no really, there’s a change in there that you can’t see.

So let’s summarize these results in a table.

My take-away from this table is that if you do your work with any of these systems, you need to pay close attention when dealing with files that contain characters outside the 7-bit ASCII set because changes to encoding or the presence of a BOM can be hard to spot, or even become outright invisible, even though it drastically changes what the contents of the file mean.

The post How various git diff viewers represent file encoding changes in pull requests appeared first on The Old New Thing.

Car horns – or at least some manner of loud noisemaker – have been around a long, long time. If we consider trains to be an early form of automobile (and, really, they are) then we’re talking 1830s or so. Even if we’re much more conservative and limit the discussion to automobiles as we know them today, we’re still talking over a century of cars with horns. And, in all that time, there have been remarkably few innovations in the Voice of the Car, with the only real recent new application of the horn –outside of its usual signal and communications role – being the honks given when locking or unlocking the car via a remote locking system. And maybe the loud, panicked honks of a car alarm. Well, at least that’s what I thought until I heard – quite belatedly – about how Nissan has re-purposed the horn for a really clever function.

There’s a lot I really like about this system I’m going to tell you about, mostly because it manages to solve a tricky and interesting problem using hardware that’s already on the car in the first place, which is one of the best solutions to a problem like this, where it requires no new equipment. I should just tell you what it is already; I’m not sure why I’m dragging this out.

It’s Nissan’s Easy-Fill Tire Alert system! Sure, the name sounds like a marketing team used to coming up with names for oven cleaners came up with it, but the system is just so damn clever I want to spit. Here, look, this is how it works:

Hot pickles, it’s so flapjacking simple and clever! I love it, I really do. This is something that actually takes a real, genuine, actually-occurs-in-reality problem and applies a clever re-use of the car’s equipment to solve it. When you’re inflating your tires, it’s an ass-pain to constantly have to stop and check the pressures. But here, the car just tells you, and it does so in an un-ignorable and wildly simple way: it honks the horn. Oh, and flashes the lights, too. And, if you over-inflate, it honks twice.

You’ll be able to know you hit the right psi even in a loud, noisy gas station, even if you’re only half-paying attention because you’re scrolling Instagram with your other hand or whatever.

Nissan made a little video that addresses just this idea by comparing their novel system to the old, crude way that we used to inflate tires, like filthy animals, and I do like how they had the other guy in the barely-disguised Honda wear a helmet, just for fun:

It’s brilliant. And yeah, this has been around since, holy crap, 2013? Man, I have been sleeping on this.

But I don’t care; a good idea is a good idea, and I feel in this age of subscription heated seats and glove boxes that open from touchscreens it’s more important than ever to call out genuinely good ideas. And, I feel like with how relatively stagnant Nissan has been for the past decade or so, this novel and clever innovation is even more worthy of recognition.

Also, the statistic shown in that video is sort of alarming:

Is this true? Or is this just a psy-op from Big Compressed Air, trying to get us to constantly inflate our tires, kind of like how the Mattress-Industrial-Complex says you need to change your mattress every 72 hours or something like that. I checked, though, and it seems to be true – an NHTSA study found that only 19% of drivers (which is about 1 out of 5) have properly inflated tires! So, really, anything that helps fix that is good.

I’m not sure how I’ve slept on this clever innovation, but I’m trying to make up for it now. And other carmakers have jumped on this as well, like Honda, who calls it Tire Fill Assist:

…though Honda’s system seems to use a dedicated beeper instead of just re-purposing the horn.

This is clever, and I’m just happy to see it. I hope it spreads to all cars! I don’t know if that means Nissan gets a bunch of licensing money, but if so, sure, why not? They can use it.

Here’s How Dirt Cheap Tire Pressure Gauges Compare To Expensive Ones: Project Farm

How To Find Out The Right Tire Air Pressure For Your Car In 30 Seconds

Why Don’t We Make Tires Like This Anymore, And Weird NASCAR Trivia: Cold Start

The post Nissan Has Somehow Come Up With One Of The Best Uses Of A Horn On A Modern Car appeared first on The Autopian.

Ahead of a major crackdown on illegal cryptocurrency mines in Russia next year, a power provider in Siberia has been fined for illegally leasing state land that's supposed to be used only for public utilities to an illegal mining operation.

In a social media post translated by Ars, the Irkutsk Region Prosecutor-General’s Office explained that the power provider was fined more than 330,000 rubles (about $3,000) for the improper land use. Local prosecutors will also pursue an administrative case against the power provider, the office said.

Crypto mining is popular in Siberia because of low operating costs, Crypto News noted, due to the cool temperatures and cheap power supply. But many in Siberia have blamed crypto miners for power outages and grid instability that can cause significant harms during winter months.

Russia's relationship with crypto mining is complicated. Last year, Russia became the world's second largest Bitcoin miner, the Moscow Times reported, mining more than $3 billion in bitcoin in 2023. Embracing cryptocurrency, the Russian government quickly moved to make crypto mining legal and approved its use in international trade, after reportedly generating approximately $550 million in tax revenue.

But Russia presumably gets no taxes on illegal crypto mining, and power outages can be costly for everyone in a region. So next year, Russia will ban crypto mining in 10 regions for six years and place seasonal restrictions that would disrupt some crypto mining operations during the coldest winter months in regions like Irkutsk, CoinTelegraph reported.

Illegal mining is still reportedly thriving in Irkutsk, though, despite the government's attempts to shut down secret farms. To deter any illegal crypto mining disrupting power grids last year, authorities seized hundreds of crypto mining rigs in Irkutsk, Crypto News reported.

In July, Russian president Vladimir Putin linked blackouts to illegal crypto mines, warning that crypto mining currently consumes "almost 1.5 percent of Russia’s total electricity consumption," but "the figure continues to go up," the Moscow Times reported. And in September, Reuters reported that illegal mines were literally going underground to avoid detection as Russia's crackdown continues.

Even though illegal mines are seemingly common in parts of Siberia and increasingly operating out of the public eye, finding an illegal mine hidden on state land controlled by an electrical utility was probably surprising to officials.

The power provider was not named in the announcement, and there are several in the region, so it's not currently clear which one made the controversial decision to lease state land to an illegal mining operation.

Read full article

Comments