Enlarge

On Wednesday, OpenAI announced DALL-E 3, the latest version of its AI image-synthesis model that features full integration with ChatGPT. DALL-E 3 renders images by closely following complex descriptions and handling in-image text generation (such as labels and signs), which challenged earlier models. Currently in research preview, it will be available to ChatGPT Plus and Enterprise customers in early October.

Like its predecessor, DALLE-3 is a text-to-image generator that creates novel images based on written descriptions called prompts. Although OpenAI released no technical details about DALL-E 3, the AI model at the heart of previous versions of DALL-E was trained on millions of images created by human artists and photographers, some of them licensed from stock websites like Shutterstock. It's likely DALL-E 3 follows this same formula, but with new training techniques and more computational training time.

Judging by the samples provided by OpenAI on its promotional blog, DALL-E 3 appears to be a radically more capable image-synthesis model than anything else available in terms of following prompts. While OpenAI's examples have been cherry-picked for their effectiveness, they appear to follow the prompt instructions faithfully and convincingly render objects with minimal deformations versus existing models. Compared to DALL-E 2, OpenAI says that DALL-E 3 refines small details like hands more effectively, creating engaging images by default with "no hacks or prompt engineering required."

Enlarge (credit: Getty | picture alliance)

Americans will again have an opportunity to receive free at-home COVID-19 rapid tests from the US government, with orders beginning next Monday, September 25, the Biden administration announced Wednesday.

Households will be eligible to receive four free rapid tests that will "detect the currently circulating COVID-19 variant," the Department of Health and Human Services said in an announcement. The tests, available next week via COVIDTests.gov and expected to start shipping on October 2, are meant to help Americans detect COVID-19 and keep from spreading it for the rest of the year—especially during holiday gatherings.

"At this point, our focus is getting through the holidays and making sure folks can take a test if they’re going to see Grandma for Thanksgiving,” Dawn O’Connell, assistant secretary for preparedness and response at the HHS, told the Associated Press.

Adult female worm of Angiostrongylus cantonensis recovered from rat lungs with characteristic barber-pole appearance (anterior end of worm is to the top). Scale bar = 1 mm. (credit: Lindo et al.)

The dreaded rat lungworm—a parasite with a penchant for rats and slugs that occasionally finds itself rambling and writhing in human brains—has firmly established itself in the Southeast US and will likely continue its rapid invasion, a study published this week suggests.

The study involved small-scale surveillance of dead rats in the Atlanta zoo. Between 2019 and 2022, researchers continually turned up evidence of the worm. In all, the study identified seven out of 33 collected rats (21 percent) with evidence of a rat lungworm infection. The infected animals were spread throughout the study's time frame, all in different months, with one in 2019, three in 2021, and three in 2022, indicating sustained transmission.

Although small, the study "suggests that the zoonotic parasite was introduced to and has become established in a new area of the southeastern United States," the study's authors, led by researchers at the University of Georgia College of Veterinary Medicine, concluded. The study was published Wednesday in the journal Emerging Infectious Diseases.

Enlarge / Stoke Space's Hopper 2 takes to the skies on Sunday in Moses Lake, Washington. (credit: Stoke Space)

At an airfield in Eastern Washington on Sunday, the small launch company Stoke Space flew its upper stage for the first time.

The flight was, admittedly, rather modest. The second-stage rocket only ascended to about 30 feet (9 meters) and traveled just several feet down range. The entire flight was over in 15 seconds.

And yet this was a momentous step for Stoke Space, which is less than 4 years old and has only about 90 employees. The test successfully demonstrated the performance of the company's oxygen-hydrogen engine, which is based on a ring of 30 thrusters; the ability to throttle this engine and its thrust vector control system; as well as the vehicle's avionics, software, and ground systems.

Enlarge / Rocket Lab's Electron launch vehicle lifts off Tuesday from New Zealand on an ill-fated mission. (credit: Rocket Lab)

Rocket Lab's string of 20 consecutive successful launches ended Tuesday when the company's Electron rocket failed to deliver a small commercial radar imaging satellite into orbit.

The problem occurred on the upper stage of the Electron rocket about two and a half minutes after liftoff from Mahia Peninsula in New Zealand. This was the fourth time a Rocket Lab mission has failed in 41 flights.

In a statement, Rocket Lab said it is "working closely" with the Federal Aviation Administration and supporting agencies as the company begins an investigation into the cause of the failure. While Rocket Lab launches most of its missions from New Zealand, the company is headquartered in the United States, giving the FAA regulatory oversight authority over failure investigations.

Enlarge (credit: Getty Images)

Google DeepMind researchers recently developed a technique to improve math ability in AI language models like ChatGPT by using other AI models to improve prompting—the written instructions that tell the AI model what to do. It found that using human-style encouragement improved math skills dramatically, in line with earlier results.

In a paper called "Large Language Models as Optimizers" listed this month on arXiv, DeepMind scientists introduced Optimization by PROmpting (OPRO), a method to improve the performance of large language models (LLMs) such as OpenAI’s ChatGPT and Google’s PaLM 2. This new approach sidesteps the limitations of traditional math-based optimizers by using natural language to guide LLMs in problem-solving. "Natural language" is a fancy way of saying everyday human speech.

"Instead of formally defining the optimization problem and deriving the update step with a programmed solver," the researchers write, "we describe the optimization problem in natural language, then instruct the LLM to iteratively generate new solutions based on the problem description and the previously found solutions."