Peter's Quasi-Daily Blah-g

The AI landscape has started to move very, very fast: consumer-facing tools such as Midjourney and ChatGPT are now able to produce incredible image and text results in seconds based on natural language prompts, and we’re seeing them get deployed everywhere from web search to children’s books.

However, these AI applications are being turned to more nefarious uses, including spreading malware. Take the traditional scam email, for example: It’s usually littered with obvious mistakes in its grammar and spelling—mistakes that the latest group of AI models don’t make, as noted in a recent advisory report from Europol.

Think about it: A lot of phishing attacks and other security threats rely on social engineering, duping users into revealing passwords, financial information, or other sensitive data. The persuasive, authentic-sounding text required for these scams can now be pumped out quite easily, with no human effort required, and endlessly tweaked and refined for specific audiences.

In the case of ChatGPT, it’s important to note first that developer OpenAI has built safeguards into it. Ask it to “write malware” or a “phishing email” and  it will tell you that it’s “programmed to follow strict ethical guidelines that prohibit me from engaging in any malicious activities, including writing or assisting with the creation of malware.”

ChatGPT won’t code malware for you, but it’s polite about it.

OpenAI via David Nield

However, these protections aren’t too difficult to get around: ChatGPT can certainly code, and it can certainly compose emails. Even if it doesn’t know it’s writing malware, it can be prompted into producing something like it. There are already signs that cybercriminals are working to get around the safety measures that have been put in place.

We’re not particularly picking on ChatGPT here, but pointing out what’s possible once large language models (LLMs) like it are used for more sinister purposes. Indeed, it’s not too difficult to imagine criminal organizations developing their own LLMs and similar tools in order to make their scams sound more convincing. And it’s not just text either: Audio and video are more difficult to fake, but it’s happening as well.

When it comes to your boss asking for a report urgently, or company tech support telling you to install a security patch, or your bank informing you there’s a problem you need to respond to—all these potential scams rely on building up trust and sounding genuine, and that’s something AI bots are doing very well at. They can produce text, audio, and video that sounds natural and tailored to specific audiences, and they can do it quickly and constantly on demand.

The U.S. Treasury said Monday that Volkswagen, BMW, Nissan, Rivian, Hyundai and Volvo electric vehicles will lose access to a $7,500 tax credit under new battery sourcing rules.

The Treasury said the new requirements effective Tuesday will also cut by half credits for the Tesla Model 3 Standard Range Rear Wheel Drive to $3,750 but other Tesla models will retain the full $7,500 credit.

Vehicles losing credits Tuesday are the BMW 330e, BMW X5 xDrive45e, Genesis Electrified GV70, Nissan Leaf , Rivian R1S and R1T, Volkswagen ID.4 as well as the plug-in hybrid electric Audi Q5 TFSI e Quattro and plug-in hybrid (PHEV) electric Volvo S60. The Swedish carmaker is 82%-owned by China’s Zhejiang Geely Holding Group.

The rules are aimed at weaning the United States off dependence on China for EV battery supply chains and are part of President Joe Biden’s effort to make 50% of U.S. new vehicle sales by 2030 EVs or PHEVs.

Hyundai said in a statement it was committed to its long-range EV plans and that it “will utilize key provisions in the Inflation Reduction Act to accelerate the transition to electrification.”

Rivian declined to comment and the other automakers could not immediately be reached for comment.

Treasury also disclosed General Motors electric Chevrolet Bolt and Bolt EUV will qualify for the full $7,500 tax credit.

GM said earlier it expected at least some of its EVS would qualify for the $7,500 tax credit under the new rules, including the 2023 Cadillac Lyriq and forthcoming Chevrolet Equinox EV SUV and Blazer EV SUV. Treasury said all GM EVs will qualify.

Earlier, Ford Motor and Chrysler-parent Stellantis said most of their electric and PHEV models would see tax credits halved to $3,750 on April 18. Treasury confirmed the automakers’ calculations.

The rules were announced last month and mandated by Congress in August as part of the $430 billion Inflation Reduction Act (IRA).

The IRA requires 50% of the value of battery components be produced or assembled in North America to qualify for $3,750, and 40% of the value of critical minerals sourced from the United States or a free trade partner for a $3,750 credit.

The law required vehicles to be assembled in North America to qualify for any tax credits, which in August eliminated nearly 70% of eligible models and on Jan. 1 new price caps and limits on buyers income took effect.

Last week, the Environmental Protection Agency proposed new emissions rules that forecast 60% of new vehicle sales in 2030 would be EVs.

A preliminary administration analysis found nearly 65% of first quarter EV sales qualified under North American final assembly and price cap requirements; more than 90% of those previously eligible first quarter sales remain eligible for at least a $3,750 credit.

Treasury in December said EVs ineligible for the $7,500 consumer tax credit could qualify for a commercial leasing $7,500 credit.

Related video:

The stunning capabilities of ChatGPT, the chatbot from startup OpenAI, has triggered a surge of new interest and investment in artificial intelligence. But late last week, OpenAI’s CEO warned that the research strategy that birthed the bot is played out. It’s unclear exactly where future advances will come from.

OpenAI has delivered a series of impressive advances in AI that works with language in recent years by taking existing machine-learning algorithms and scaling them up to previously unimagined size. GPT-4, the latest of those projects, was likely trained using trillions of words of text and many thousands of powerful computer chips. The process cost over $100 million.

But the company’s CEO, Sam Altman, says further progress will not come from making models bigger. “I think we’re at the end of the era where it’s going to be these, like, giant, giant models,” he told an audience at an event held at MIT late last week. “We’ll make them better in other ways.”

Altman’s declaration suggests an unexpected twist in the race to develop and deploy new AI algorithms. Since OpenAI launched ChatGPT in November, Microsoft has used the underlying technology to add a chatbot to its Bing search engine, and Google has launched a rival chatbot called Bard. Many people have rushed to experiment with using the new breed of chatbot to help with work or personal tasks.

Meanwhile, numerous well-funded startups, including Anthropic, AI21, Cohere, and Character.AI, are throwing enormous resources into building ever larger algorithms in an effort to catch up with OpenAI’s technology. The initial version of ChatGPT was based on a slightly upgraded version of GPT-3, but users can now also access a version powered by the more capable GPT-4.

Altman’s statement suggests that GPT-4 could be the last major advance to emerge from OpenAI’s strategy of making the models bigger and feeding them more data. He did not say what kind of research strategies or techniques might take its place. In the paper describing GPT-4, OpenAI says its estimates suggest diminishing returns on scaling up model size. Altman said there are also physical limits to how many data centers the company can build and how quickly it can build them.

Nick Frosst, a cofounder at Cohere who previously worked on AI at Google, says Altman’s feeling that going bigger will not work indefinitely rings true. He, too, believes that progress on transformers, the type of machine learning model at the heart of GPT-4 and its rivals, lies beyond scaling. “There are lots of ways of making transformers way, way better and more useful, and lots of them don’t involve adding parameters to the model,” he says. Frosst says that new AI model designs, or architectures, and further tuning based on human feedback are promising directions that many researchers are already exploring.

Each version of OpenAI’s influential family of language algorithms consists of an artificial neural network, software loosely inspired by the way neurons work together, which is trained to predict the words that should follow a given string of text.

Rent prices rose by an average of 14% last year, and it doesn’t look like those heart-clutching price hikes are going to stop any time soon. For anyone wondering exactly how much they’re getting ripped off in this rental hellscape, there are tools like RentHop RentCompare, which is specifically designed to help renters in New York City figure out if they’re overpaying or underpaying for rent. While the app is specific to NYC homes right now, the price tool is still a great way to explore how much different neighborhoods cost and help you see where you stand in the current rental market.

Here’s how you can use this rent compare tool, no matter where you live.

How to use RentCompare

RentCompare is an interactive tool that allows renters to compare a rental price across the average rent for an apartment’s size and neighborhood. On renthop.com, you enter your email and your neighborhood, current rent, apartment size, and amenities. Then, RentCompare reveals how it compares to similar apartments in the area.

The goal is to “put the power back in renters’ hands” by showing them whether their current rental price is below the average for apartments in the neighborhood with the same bedroom size. Of course, you may pay more than the average renter if you’re also paying for amenities like a doorman, roof deck, or in-unit washer/dryer (my greatest dream).

Consider what these famous fictional New Yorkers were “paying”

If you’re looking for an apartment in NYC, you can use a rental listings site like Zillow and compare the listed prices to comparable apartments listed on RentHop’s RentCompare. And if you live somewhere that is not NYC, you can use Zillow to compare rent prices to what is most likely a much more affordable city. Or, you can still play around with tool—just pretend to be some of your favorite famous New Yorkers.

A few years back, Insider got to the bottom of how much NYC’s most famous fictional apartments actually cost. Plug in some of the truly exorbitant apartment prices below and see whether or not Seinfeld was getting a fair deal. (What’s the deal with getting a fair deal, am I right?) (Spoiler: He was getting a great deal, considering the fun fact that Seinfeld’s TV apartment is physically impossible). Mess around with these wild home prices:

• Friends (90 Bedford Street, West Village): $2 million
• Sex And The City (66 Perry Street, West Village): $600,000
• Seinfeld (129 West 81st Street, Upper West Side): $900,000
• Big Daddy (16 Mercer Street, SoHo): $2.5 million
• The Cosby Show (Brooklyn Heights): $11 million
• Succession (specifically Kendall Roy’s ex-wife) (Tribeca’s Woolworth Tower Residences, Tribeca): $23.355 million (This one was pulled from Architectural Digest.)

Keep in mind that rent varies across the city, while most fictional New Yorkers seem to get the luxury of living on the Upper West Side. Last year, GBSN Research found that the median rent is $3,550 in Manhattan, $2,850 in Brooklyn, and $2,400 in Queens.

An entry into the World Photography Organization’s Sony World Photography Awards has brought new attention to the conversation around AI-generated art. After artist Boris Eldagsen entered his AI-generated piece “The Electrician” into the art contest and won, he opted to turn down the prize.

“The Electrician” features two women—one leaning her face on the other’s back—in a grainy sepia tone with streaks of light that recall the “spirit photography” of the early 20th century. It’s a beautiful work of art that showcases clear talent from the artist, except the work was created in collaboration with an AI used by Eldagsen, who refers to himself as a “photomedia artist” on his website. Eldagsen entered the Open Category at the Sony World Photography Awards according to his website and won with “The Electrician.” As a winner, Eldagsen would have received photography equipment from Sony, inclusion in a World Photography Organization’s book and exhibit, as well as plenty of exposure for his career.

“I have been photographing since 1989, been a photomedia artist since 2000. After two decades of photography, my artistic focus has shifted to exploring the creative possibilities of AI generators,” Eldagsen wrote on his website. “The work SWPA has chosen is the result of a complex interplay of prompt engineering, inpainting and outpainting that draws on my wealth of photographic knowledge. For me, working with AI image generators is a co-creation, in which I am the director. It is not about pressing a button—and done it is.”

As Peta Pixel reports, the World Photography Organization announced the awards winners on March 14, with Eldagsen being among the finalists. A month later, on April 13, Eldagsen announced on his website that he would not accept the award. As Eldagsen writes in his announcement, he applied to the contest as a “cheeky monkey” looking to test if “competitions are prepared for AI images. They are not.” The World Photography Organization did not immediately return Gizmodo’s request for comment on Eldagsen’s piece or his decision to refuse the prize.

The conversation around the applications of AI has seemingly come to a head in recent months, and Eldagsen was not the first artist to win a competition with AI. In August of last year, Jason Allen submitted his piece, titled Théåtre D’opéra Spatial, to the Colorado State Fair Fine Arts competition, and he won. In that case, the artist was happy to accept the accolades.

As the AI boom continues, there has been a substantial rise in AI-generated art, with Runway’s Gen-2, and the infamous DALL-E 2 from OpenAI demonstrating how easy it can be for a computer to create video and images from text. The issue with most of these generators is their clear tendency to lift pre-existing (often copyrighted) material from other sources, only to Frankenstein it into an original piece of art.

Want to know more about AI, chatbots, and the future of machine learning? Check out our full coverage of artificial intelligence, or browse our guides to The Best Free AI Art Generators, The Best ChatGPT Alternatives, and Everything We Know About OpenAI’s ChatGPT.