GPT-4 is here, and we’re one step closer to a world where AI ability outstrips our ability to use it wisely

Prior to the GPT-4 benchmarks being released, a good number of researchers were saying all the AI excitement was mostly hype, and that we are nowhere near these LLM approaching natural human language capabilities in anything but the most simple applications.

After GPT-4 I think there are more people who worried about where this is going faster than I think a lot of people imagined it will be going. As processing speeds increase and the sheer volume of data ingested starts to approach heretofore unseen levels, I think we’re really not ready for this. We can’t even, as a society, manage to get a handle on legislating and adjudicating, much less being able to predict where AI is taking us and how we should (or should not) use it.

Charlie Warzel (“Galaxy Brain”), one of the only truly reliably trenchant, useful and interesting writers in the often disappointingly mediocre Atlantic Magazine, has some thoughts from himself and others:

There’s always been tension in the field of AI—in some ways, our confused moment is really nothing new. Computer scientists have long held that we can build truly intelligent machines, and that such a future is around the corner. In the 1960s, the Nobel laureate Herbert Simon predicted that “machines will be capable, within 20 years, of doing any work that a man can do.” Such overconfidence has given cynics reason to write off AI pontificators as the computer scientists who cried sentience!

Melanie Mitchell, a professor at the Santa Fe Institute who has been researching the field of artificial intelligence for decades, told me that this question—whether AI could ever approach something like human understanding—is a central disagreement among people who study this stuff. “Some extremely prominent people who are researchers are saying these machines maybe have the beginnings of consciousness and understanding of language, while the other extreme is that this is a bunch of blurry JPEGs and these models are merely stochastic parrots,” she said, referencing a term coined by the linguist and AI critic Emily M. Bender to describe how LLMs stitch together words based on probabilities and without any understanding. Most important, a stochastic parrot does not understand meaning. “It’s so hard to contextualize, because this is a phenomenon where the experts themselves can’t agree,” Mitchell said.

One of her recent papers illustrates that disagreement. She cites a survey from last year that asked 480 natural-language researchers if they believed that “some generative model trained only on text, given enough data and computational resources, could understand natural language in some non-trivial sense.” Fifty-one percent of respondents agreed and 49 percent disagreed. This division makes evaluating large language models tricky. GPT-4’s marketing centers on its ability to perform exceptionally on a suite of standardized tests, but, as Mitchell has written, “when applying tests designed for humans to LLMs, interpreting the results can rely on assumptions about human cognition that may not be true at all for these models.” It’s possible, she argues, that the performance benchmarks for these LLMs are not adequate and that new ones are needed.

There are plenty of reasons for all of these splits, but one that sticks with me is that understanding why a large language model like the one powering ChatGPT arrived at a particular inference is difficult, if not impossible. Engineers know what data sets an AI is trained on and can fine-tune the model by adjusting how different factors are weighted. Safety consultants can create parameters and guardrails for systems to make sure that, say, the model doesn’t help somebody plan an effective school shooting or give a recipe to build a chemical weapon. But, according to experts, to actually parse why a program generated a specific result is a bit like trying to understand the intricacies of human cognition: Where does a given thought in your head come from?

I’ve come to the conclusion that, as dangerous as AI could become, the most frightening thing that can happen in the here and now is how people will anthropomorphize it and believe what it spits out even when it is wrong. In a world where a majority of people are scientifically and civically illiterate, having something that many people believe is sentient and infallible is a danger that is on our doorstep.

All an evil someone with sufficient AI computer knowledge and coding skills need do is find a way to exploit those two things; trust in AI infallibility and the belief that what is in the all-knowing computer holds your interests and well-being paramount.

Capitalism will find a way to exploit that long before any computer reaches sentience.

Imagine being trapped in a foreign dictatorship while being forced to pay a business debt that isn’t yours

This story is wild:

It has been nearly five years since police here told Henry Cai, a U.S. citizen from California, that he couldn’t leave China.

Just before Christmas 2017, he was stopped at the airport at the end of a business trip. Mr. Cai later learned somebody was trying to force him to pay an outstanding debt of several million dollars owed by a Beijing company where he was a director and shareholder.

He thought it was a misunderstanding and expected it to be sorted out quickly. And yet here he remains, stuck in China, the target of a form of Chinese justice known as an exit ban.

His is believed to be the longest-running case of such legal purgatory for an American businessman. Now 61 years old, Mr. Cai has wrangled with China’s judicial bureaucracy, tested the limits of U.S. diplomacy and depleted his savings.

In an interview with The Wall Street Journal he said he fears deteriorating U.S.-China relations—which are in the spotlight with the first meeting between President Biden and Chinese leader Xi Jinping as heads of state on Monday—have worsened his quandary.

The U.S. has accused China of using exit bans on Americans and other foreigners “without fair and transparent process under the law.” Diplomats say Americans trapped in legal jeopardy abroad increasingly occupy their time.

Mr. Cai hasn’t been charged with a crime. Instead, court records outline a financial dispute between business partners that has included a police investigation. Mr. Cai said he is being squeezed to pay a debt that isn’t his.

You can read the rest of the Wall Street Journal article by James T. Areddya and Brian Spegele at this link.

Henry Cai, trapped in China. (WSJ photo.)

“This is a Bernie Madoff-size scam for the sneaker market”

There are so many telltale signs of the corroding effects of capitalism on society that I’d hardly know where to start (or end) in presenting the evidence.

But this Wall Street Journal article about the reseller market in high-end sneakers is right up there.

Mr. Malekzadeh’s apparent success afforded him the kind of insouciant, gold-plated lifestyle that luxury sneakers are thought to reflect. On Instagram, the 39-year-old showed off his Ferraris and a six-figure Girard-Perregaux watch next to a hamburger. He also posted shots of himself riding a $29,000 Louis Vuitton bicycle inside his million-dollar home in Eugene, Ore.

The business, in real life, was collapsing under the weight of unfulfilled orders, late payments and customer complaints. In May, Mr. Malekzadeh’s fiancée—also the company’s finance chief—pushed for both of them to come clean, according to people familiar with the situation.

Federal prosecutors a few months later charged the couple with bank fraud and Mr. Malekzadeh with wire fraud and money laundering. Customers claim they paid millions of dollars for shoes that never arrived. A court-appointed receiver is sorting out the remaining inventory of the entrepreneur’s company, Zadeh Kicks.

Early last year, Mr. Malekzadeh collected orders for about 600,000 pairs of Air Jordan 11 Cool Grey sneakers months before they hit stores, netting over $70 million, according to prosecutors. He priced the sneakers between $115 and $200 a pair, cheaper than their expected retail price of around $225, prosecutors said.

Mr. Malekzadeh was able to get only 6,000 pairs.

Prosecutors allege he collected preorder funds from customers while knowing he couldn’t fill all the orders. Since at least 2020, he spent more than $10 million of the company’s preorder proceeds on luxury goods, including watches, furs and handbags, they said. In a seizure warrant affidavit, federal authorities allege the couple also used customer money to help make a down payment on a house and complete about $600,000 of work to remodel it.

Seriously, man, Yeezy and Travis Scott are both terrible human beings. The evidence has been out there for some time that they are not the kind of people you’d want in your social circle unless you are a hopeless social climber.

Yet someone people still pay thousands for sneakers with their names on them.

Of course a business like that would be ripe for con artists. I don’t feel a bit of sympathy for anyone who gets swindled in this maelstrom of greed and status-seeking. You’re stupid enough to pay more than $100 for a pair of sneakers, you deserve to get cheated.

Malekzadeh in his empire that was apparently largely smoke and mirrors.

Just one measure of the fucked-up society capitalism is creating

The picture below is from yesterday’s (11/1/22) New York Times in an article titled “What Does the End of Yeezy Mean for the Sneakerverse?”

I think Yeezy, despite whatever mental health issues he has, is a terrible person who should be de-platformed in whatever ways capitalism now allows.

That is the market speaking, not cancel culture. Whatever could be more free than the free market deciding en masse that Kanye West is too crazy and bigoted to be trusted with a company’s entire product line?

But the larger question is this:

Why are we worried about drag queen story hours — a tiny problem, no matter how you feel about them — when we are creating a culture where sneakers are bought and sold for tens of thousands of dollars like commodities? Is this not a sign of a much sicker society than any man in women’s clothing reading “Cat In The Hat” books?

FAA rejects airline’s bid to cut pilot training hours in half

A major regional airline’s push to have less training for pilots is rejected by the FAA.

Aviation-safety regulators rejected a proposal by a regional airline seeking to reduce the number of hours that some co-pilots need to begin flying passengers.

Indianapolis-based Republic Airways Holdings Inc. had asked the Federal Aviation Administration to allow pilots who go through a special program at the airline’s training academy to begin flying on a restricted license after 750 hours of training—half what is generally required.

The request came as regional airlines such as Republic say they are facing a shortage of pilots that has strained their ability to fly to small cities around the country. Republic operates flights for United Airlines Holdings Inc., American Airlines Group Inc. and Delta Air Lines Inc.

The FAA said in a letter to Republic on Monday that it didn’t agree that Republic’s plan served the public interest and doesn’t believe the airline’s plan would help address a “perceived pilot shortage.” The agency said granting an exemption to Republic could open the door to similar requests from other airlines.

The decision underscores the dilemma facing regional airlines, which are generally smaller carriers that play an outsize role in U.S. air travel, operating over 40% of U.S. passenger flights.

Republic’s request reignited an industry debate about a federal rule that requires aspiring U.S. airline pilots to have at least 1,500 hours of flying experience to qualify to be a first officer at an airline, unless they are former military pilots or graduates of colleges and universities with professional aviation programs.

That requirement, dubbed the 1,500-hour rule, was put in place in 2013 after a fatal plane crash in 2009 near Buffalo, N.Y., which investigators blamed on a tired crew that didn’t properly react to stall warnings. The Air Line Pilots Association (APLA), a pilots union, opposed Republic’s request and disputes that there is a pilots shortage.

If I believed anyone, I’d believe the APLA. And I’ll be avoiding any flights on Republic henceforth. If they’re trying to jeopardize passenger and pilot safety just to increase their bottom line by hiring less-experienced pilots, who knows what other ways they are cutting corners?

Ga. governor signs off on Hyundai’s fleecing of his state’s taxpayers

Now that incumbent Georgia Gov. Brian Kemp overwhelmingly won his primary over Sonny Perdue — thus helping to prove that Trump is not all-powerful — we can go back to loathing Kemp.

He’s still a Trump toadie, despite having political courage and principles just that one time when he stood up to Stop The Steal nuts. (It doesn’t take all that much courage to look at someone and say, “This is not what the law says.” The only reason it needs to be celebrated at all is because it stands out so much from normal everyday GOP corruption and narcissism. It’s not that Kemp is that principled, it’s that people surrounding him in his party are so devoid of scruples.)

To that end, it’s important that everyone point out that Kemp’s deal to build a Hyundai assembly plant in Ga. is a massive give-away to Hyundai that will create jobs at a shocking cost of $228,000 per job. And it will rob state public works programs to pay for making Hyundai richer.

For that kind of a shitty deal, you have to wonder who’s getting paid off where.

The folks at ITPI are on the case:

Take Georgia’s new big, shiny $1.8 billion factory deal with South Korean automaker Hyundai. The state’s republican governor Brian Kemp is making it sound like a win-win for all involved.

“We are proud to welcome Hyundai Motor Group to Georgia as we forge an innovative future together,” Kemp said in a press release. “We will continue working to make Georgia the premier destination for quality companies who are creating the jobs of today, tomorrow, and beyond.”

But when you actually look at the terms of the deal—which is the largest subsidy package for an automotive plant ever in the U.S.—your head can’t help but hurt. There’s a reason Georgia officials wouldn’t reveal what incentives Hyundai had been promised until after the agreement was signed.

Here they are:

  • Local governments are giving Hyundai more than $472 million in property tax breaks.
  • The company will also receive more than $212 million in state corporate income tax credits. (Get this: If Hyundai doesn’t end up owing that much in state income tax, Georgia will instead give the company personal income taxes collected from the company’s workers.)
  • The state and local governments spent $86 million to purchase the plant site
  • Georgia will spend $200 million on road construction and improvements, plus $50 million more to help fund construction, machinery, and equipment.
  • Sales tax exemptions on construction materials and machinery expenses are estimated to cost $396 million.

All in all, Georgia and four counties will be giving Hyundai about $228,000 per job created.

You can read the rest of the article at this link.

Ga. Governor Brian Kemp signs-off on a plan to let Hyundai raid his state’s treasury .

FCC orders phone carriers to stop carrying some auto warranty robocalls

I’m not sure how (and whether) this makes phone carriers liable except insofar as making reports about what steps they are taking to stop auto warranty robocalls, but it’s a step in the right direction.

The FCC’s Robocall Response Team today announced that the Enforcement Bureau has ordered phone companies to stop carrying traffic regarding a known robocall scam marketing auto warranties.  The calls are coming from Roy Cox, Jr., Aaron Michael Jones, their Sumco Panama companies, and international associates.  Building on FCC action earlier this month, all U.S. voice service providers must now take all necessary steps to avoid carrying this robocall traffic.  This operation is also the target of an ongoing investigation by the FCC’s Enforcement Bureau and a lawsuit by the Ohio Attorney General.

FCC Chairwoman Jessica Rosenworcel:

“We are not going to tolerate robocall scammers or those that help make their scams possible.  Consumers are out of patience and I’m right there with them.”

What’s New:

The Enforcement Bureau has ordered all U.S. voice service providers to take all necessary steps to avoid carrying robocall traffic from the Cox/Jones/Sumco Panama operation.  Today’s order followed a Public Notice that warned providers of this concerning flood of robocalls.  The notice had authorized providers to cut off the traffic and today’s order requires that they do so.  If they do not, they must regularly report to the FCC of the steps they have taken to mitigate the traffic.

“Now that U.S. voice service providers know the individuals and entities associated with this scheme, the Enforcement Bureau will closely monitor voice service providers’ compliance with this order and take appropriate enforcement action as necessary,” said Acting FCC Enforcement Bureau Chief Loyaan A. Egal.

Hooray for the Biden administration if this works.

You can read the rest of the FCC’s statements about the change here.

Bitcoin industry claims it is helping to prevent Texas power outages

Or so they say:

As a blistering heat wave smothered Texas this week and taxed the power grid with record-high demand, bitcoin mining operators in the state shut down their electricity-guzzling machines.

Complying with requests from the Electric Reliability Council of Texas — the grid operator that asked businesses and residents to voluntarily conserve electricity during the heat wave — nearly all industrial-scale mining in the state reportedly powered down, according to the Texas Blockchain Council, an industry association.

Cryptocurrency mining requires huge amounts of electricity, prompting concerns not only around whether Texas’s beleaguered grid can keep up with skyrocketing demand as more miners are expected to flock to the state, but also over the industry’s broader potential environmental impact.

“There are over 1,000 megawatts worth of bitcoin mining load that responded to ERCOT’s conservation request by turning off their machines to conserve energy for the grid,” Lee Bratcher, president of Texas Blockchain Council, told Bloomberg News in an email.

Of course, an industry that is melting down before everyone’s eyes and taking millions of people’s life savings in those losses. has no reason at all right now to lie in ways that will take attention away from those facts.

Reporters today are often such shitheads. They “fact check” doctors who says a 10-year-old had to go to Indiana for an abortion, but take the word of an entire shady industry as gospel when they say they are helping the environment and energy loads.

And what’s even worse?

The WaPo reporter in question never once mentioned in her article that crypto was in a freefall. I think that might have something to do with the fact that those guys are willing to cut back.

Anyway, you can read the rest here.

What do you get when you mix a love of Ayn Rand and child pornography?

I subscribe to The New Yorker because at least once or twice an issue, a truly interesting and/or bizarre article makes its pages. Of course, you have to wade through endless articles with New Yorkers navel-gazing about boring New York-centric things, including things that have a knowing only-in-New-York air about them despite the fact that most of them happen in cities across America every day.

Anyway, the current issue has once of those engrossing and weird articles, titled, The Surreal Case of a CIA Hacker’s Revenge. It concerns Joshua Schulte, a CIA coder who has, to put it mildly, interpersonal issues with others. First of all, he remains a devotee of Ayn Rand well into adulthood, long after most thinking people discard her for the nut she was. But Schulte has gone all-in on the “Selfishness is Good” mantra.

He loves Ron Paul. He loves arguing about his wild libertarian beliefs. He alienates everyone around him. In other words, he’s just like all of your obsessed libertarian friends.

And he allegedly leaked CIA secrets, which has landed him in much hot water.

Here is one section of the article:

Other classmates recalled sexually inappropriate behavior. One woman told me that he had repeatedly exposed his penis to students when they were both in the junior-high band. “He would try and touch people, or get people to touch him—that was a daily occurrence,” she said. She loved music, but she was so intent on getting away from Schulte that she asked her parents to let her quit the band. She was too uncomfortable to explain to her parents exactly what had transpired. “It’s hard to put it into words,” she recalled. “You’re twelve. It’s just ‘Hey, this kid is super gross, and it makes me want to not be part of this school right now.’ ” Her parents, not grasping the gravity of what had happened, insisted that she remain in the band. “I was traumatized,” she told me. I also spoke to a friend of the woman, who remembered her recounting this behavior by Schulte at the time. A third woman told me that Schulte and some of his friends got in trouble at school after trying to stick their hands into her pants while she slept on the bus during a field trip. Schulte, she said, took revenge by sending her an AOL message loaded with a virus, destroying her computer. He boasted about the hack afterward, the woman said.

Schulte’s friend Kavi Patel acknowledged that Schulte would “draw swastikas all over the place.” He wasn’t anti-Semitic, Patel contended; he just relished getting a rise out of people. He recalled Schulte telling him, “I don’t really care one way or the other, but it’s fun to see the shock on people’s faces.” Patel was also in the junior-high band. When I asked him if he remembered Schulte exposing himself, he said that he never witnessed it, but had heard about it happening “two or three times.” According to Patel, Schulte seemed to confirm it to him on one occasion: “I was, like, ‘Dude, did you do this?’ And he was, like, ‘Heh, heh.’ ” Patel added, “It’s not something that’s out of his character. At all.” (Presented with these allegations, several attorneys who have represented Schulte had no comment. Deanna recalled learning that Joshua had drawn a swastika in his notes for a lesson on the Second World War, but she and Roger said that they were not aware of other incidents involving swastikas or the junior-high band. They dispute the classmate’s recollection of the incident on the school bus.)

When Schulte was in college, he argued on his blog that pornography is a form of free expression which “is not degrading to women” and “does not incite violence.” He went on, “Porn stars obviously enjoy what they do, and they make quite a bit of money off it.” Of course, some women are coerced into pornography, and if you mistake the simulated enjoyment in a porn performance for the real thing then you don’t understand much about the industry. But more to the point: child pornography is not free expression; it’s a crime. After Schulte realized that the illicit archive had been discovered, he claimed that the collection—more than ten thousand images and videos—didn’t belong to him. In college, he had maintained a server on which friends and acquaintances could store whatever they wanted. Unbeknownst to him, he contended, people had used the server to hide contraband. He “had so many people accessing it he didn’t care what people put on it,” Roger Schulte told the Times.

But, according to the F.B.I., as agents gathered more evidence they unearthed chat logs in which Schulte conversed about child pornography with fellow-enthusiasts. “Where does one get kiddie porn anyways?” Schulte asked, in a 2009 exchange. This was another instance in which Schulte seemed recklessly disinclined to cover his tracks. His Google search history revealed numerous queries about images of underage sex. In the chat logs, people seeking or discussing child pornography tended to use pseudonyms. One person Schulte interacted with went by “hbp.” Another went by “Sturm.” Josh’s username was “Josh.” At one point, he volunteered to grant his new friends access to the child-porn archive on his server. He had titled it /home/josh/http/porn. Sturm, taken aback, warned Schulte to “rename these things for god’s sake.”

The article, by Patrick Radden Keefe, is so well done, and full of jaw-dropping revelations. And the article isn’t behind the New Yorker paywall.

Joshua Schulte

The drone that is keeping the Russians at bay in Ukraine

The brainchild of an MIT-educated Turkish engineer, the drone is changing the face of warfare in ways that only happen every couple of generations

The New Yorker has a fascinating article in its current edition about the Bayraktar TB2, a drone that is helping to balance the scales in Russia’s war with Ukraine:

A video posted toward the end of February on the Facebook page of Valerii Zaluzhnyi, the commander-in-chief of Ukraine’s armed forces, showed grainy aerial footage of a Russian military convoy approaching the city of Kherson. Russia had invaded Ukraine several days earlier, and Kherson, a shipbuilding hub at the mouth of the Dnieper River, was an important strategic site. At the center of the screen, a targeting system locked onto a vehicle in the middle of the convoy; seconds later, the vehicle exploded, and a tower of burning fuel rose into the sky. “Behold the work of our life-giving Bayraktar!” Zaluzhnyi’s translated caption read. “Welcome to Hell!”

Because it is a fraction of the cost of Israeli and American payload-carrying drones, it’s proving to be popular:

In April, 2016, the TB2 scored its first confirmed kill. Since then, it has been sold to at least thirteen countries, bringing the tactic of the precision air strike to the developing world and reversing the course of several wars. In 2020, in the conflict between Azerbaijan and Armenia over the enclave of Nagorno-Karabakh, Azerbaijan’s dictatorial leader, Ilham Aliyev, used the TB2 to target vehicles and troops, then displayed footage of the strikes on digital billboards in the capital city of Baku.

The TB2 has now carried out more than eight hundred strikes, in conflicts from North Africa to the Caucasus. The bombs it carries can adjust their trajectories in midair, and are so accurate that they can be delivered into an infantry trench. Military analysts had previously assumed that slow, low-flying drones would be of little use in conventional combat, but the TB2 can take out the anti-aircraft systems that are designed to destroy it. “This enabled a fairly significant operational revolution in how wars are being fought right now,” Rich Outzen, a former State Department specialist on Turkey, told me. “This probably happens once every thirty or forty years.”

From an engineering standpoint this is all very interesting. From the standpoint of guerilla warfare and terrorism, I’m thinking it’s only a matter of time until the technology will advance to the point that civilians around the world will be targeted in non-combat ways we’ve not yet imagined. I’ve been reading predictions along those lines for years. It’s scary how close we’ve come to reality.

If I were Armenian? Well….

The Bayraktar TB2 on a runway. (Photo: Wikimedia Commons)