Cosmic Log
Astronomers have used machine learning to sharpen up the Event Horizon Telescope’s first picture of a black hole — an exercise that demonstrates the value of artificial intelligence for fine-tuning cosmic observations.
The image should guide scientists as they test their hypotheses about the behavior of black holes, and about the gravitational rules of the road under extreme conditions.
Can a chatbot help Pentagon planners find the satellite data they need to understand what’s happening in a global hotspot? Microsoft Azure Space recently showed the U.S. military how an application beefed up with AI could do just that.
The daylong demonstration, which was conducted for the Defense Innovation Unit’s Hybrid Space Architecture program last month, is among several space-related developments that Microsoft and its partners showcased today in advance of next week’s Space Symposium in Colorado.
Other developments include a collaboration with Ball Aerospace and Loft Federal on an experimental satellite program for the Defense Department’s Space Development Agency; a new frontier for Microsoft Azure’s partnership with Viasat; and a milestone for the Space Information Sharing and Analysis Center.
“Azure Space has been committed to enabling people to achieve more, both on and off the planet,” Stephen Kitay, senior director of Microsoft Azure Space, told me. “And this commitment encompasses not only commercial [applications], but also empowers government missions as well. Digital transformation within the government is the key to unlocking the full potential of what’s possible in space, and Microsoft is providing these technologies and solutions to government agencies alongside our partners to make this transformation possible.”
It turns out we’re not the only ones getting into fact-checking fights with Bing Chat, Microsoft’s much-vaunted AI chatbot.
Last week, GeekWire’s Todd Bishop recounted an argument with the ChatGPT-based conversational search engine over his previous reporting on Porch Group’s growth plans. Bing Chat acknowledged that it gave Bishop the wrong target date for the company’s timeline to double its value. “I hope you can forgive me,” the chatbot said.
Since then, other news reports have highlighted queries that prompted wrong and sometimes even argumentative responses from Bing Chat.
Artificial intelligence software programs are becoming shockingly adept at carrying on conversations, winning board games and generating artwork — but what about creating software programs? In a newly published paper, researchers at Google DeepMind say their AlphaCode program can keep up with the average human coder in standardized programming contests.
“This result marks the first time an artificial intelligence system has performed competitively in programming contests,” the researchers report in this week’s issue of the journal Science.
There’s no need to sound the alarm about Skynet just yet: DeepMind’s code-generating system earned an average ranking in the top 54.3% in simulated evaluations on recent programming competitions on the Codeforces platform — which is a very “average” average.
“Competitive programming is an extremely difficult challenge, and there’s a massive gap between where we are now (solving around 30% of problems in 10 submissions) and top programmers (solving >90% of problems in a single submission),” DeepMind research scientist Yujia Li, one of the Science paper’s principal authors, told me in an email. “The remaining problems are also significantly harder than the problems we’re currently solving.”
Nevertheless, the experiment points to a new frontier in AI applications. Microsoft is also exploring the frontier with a code-suggesting program called Copilot that’s offered through GitHub. Amazon has a similar software tool, called CodeWhisperer.
Oren Etzioni, the founding CEO of Seattle’s Allen Institute for Artificial Intelligence and technical director of the AI2 Incubator, told me that the newly published research highlights DeepMind’s status as a major player in the application of AI tools known as large language models, or LLMs.
“This is an impressive reminder that OpenAI and Microsoft don’t have a monopoly on the impressive feats of LLMs,” Etzioni said in an email. “Far from it, AlphaCode outperforms both GPT-3 and Microsoft’s Github Copilot.”
Are intelligent aliens living among us? A newly published novel just might lead you to think so — and in this case, the aliens aren’t visitors from another planet.
Instead, they’re octopuses, the eight-legged denizens of the deep that are celebrated in movies (including the Oscar-winning documentary “My Octopus Teacher”) and on the ice rink (thanks to the Kraken, the Seattle hockey team that’s getting set for its second NHL season.)
Ray Nayler, who wrote the novel titled “The Mountain in the Sea,” says he chose the octopus to serve as the designated alien for his science-fiction plot in part because it’s “a creature that has a structure totally different from ours, but in whom we recognize curiosity, which is what I think we find often most human in ourselves.”
Nayler doesn’t stop there: The promises and perils of artificial intelligence also figure prominently in the plot — in a way that sparks musings about how we’ll deal with AI, with kindred species on our planet, and perhaps eventually with extraterrestrial intelligence as well.
Dominic Sivitilli, a neuroscientist and astrobiologist at the University of Washington, says such musings are what led him to focus his studies on octopuses. “I suddenly had this model for what intelligence might look like, had it had a completely different evolutionary origin … possibly on another world, in another solar system,” he says. “And so they became a bit of a model to me for what extraterrestrial intelligence might end up looking like.”
Nayler and Sivitilli discuss animal intelligence, artificial intelligence and the prospects for cross-species communication in the latest episode of Fiction Science, a podcast that focuses on the intersection of science, technology, fiction and culture.
Six months after coming out of stealth mode, a Seattle AI startup called Yoodli is raising the curtain on a free website that’s designed to improve your public speaking.
And as if that’s not enough, Yoodli is throwing in some free games as well. “Think of these as Wordle, but for communication skills,” Yoodli co-founder Varun Puri said in an introductory video.
Overwatch Imaging, an Oregon venture that makes airborne imaging systems for piloted aircraft and drones, has closed on an $11.15 million investment round led by Squadra Ventures and Shield Capital.
Other participants in the round include the Portland Seed Fund and two strategic investors, L3Harris and Bridger Aerospace.
Overwatch said the fresh funding will be used to accelerate the company’s growth and enhance its automation and image analytics software. The company plans to hire 15 to 20 more engineers and business development professionals this year.
Founded in 2016, the Hood River, Ore.-based startup specializes in precision imaging systems equipped with onboard AI software. The dual-use systems can be used for defense applications as well as for civilian applications including search-and-rescue, wildfire mapping, disaster response and infrastructure inspections.
“Check my spacewalking gloves, HAL.”
The HAL 9000 computer that starred in “2001: A Space Odyssey” — and made such a mess of maintenance issues on the Discovery One spaceship — isn’t on the job on the International Space Station. But Microsoft and Hewlett Packard Enterprise are teaming up with NASA to put artificial intelligence to work on mundane orbital tasks, starting with the chore of checking spacewalkers’ gloves for wear and tear.
That’s just one of two dozen experiments in AI, cloud and edge computing that have been run on HPE’s Spaceborne Computer-2 since the hardware was sent to the space station a year ago.
“We’re bringing AI to space and empowering space developers off the planet with Azure, and it’s enabling the ability to build in the cloud and then deploy in space,” Steve Kitay, senior director of Azure Space at Microsoft, told me.
Once again, Seattle’s tech scene provides the backdrop for a high-profile movie on HBO Max — but this time, it’s serious.
Oscar-winning film director Steven Soderbergh’s tech-noir thriller, “Kimi,” echoes movies like “Rear Window” and “The Conversation” in a tale that also reflects the mind-wrenching isolation forced by the COVID-19 pandemic and the concerns raised by smart devices that are capable of tracking our every move.
Zoë Kravitz portrays an employee at a Seattle tech startup that markets a smart speaker and AI voice assistant called Kimi. The startup is gearing up for an IPO that promises a big payoff, but as Kravitz’s character works through a list of audio files that Kimi couldn’t understand, she happens upon a snippet that suggests a crime was committed. Her efforts to get to the truth spark a classic spy chase with some extra tech twists.
It’s a tale far darker than “Superintelligence,” the 2020 romantic comedy starring Melissa McCarthy as a Seattle techie and James Corden as an AI overlord.
Will “Kimi” stir up a debate over AI voice assistants? Does the movie accurately reflect the Seattle vibe? Will it generate as much buzz as Amazon’s Alexa, or will it flop as hard as the Fire Phone? The early indications are mixed: On the Rotten Tomatoes website, for example, the critical consensus is thumbs-up (89%) while the audience score is an emphatic thumbs-down (52%).
To get the verdict from ground zero, we turned to the experts who helped us sort out the fact, fiction and frivolousness in “Superintelligence”: Carissa Schoenick, director of program management and communication at Seattle’s Allen Institute for Artificial Intelligence; and Kurt Schlosser, GeekWire’s go-to guy for coverage of Seattle’s tech culture.
Can artificial intelligence write its own programs? Is there AI in your TV remote control? Researchers at Seattle’s Allen Institute for Artificial Intelligence say that knowing the right answers to such questions is an essential part of being literate in our tech-driven society — and that most of us would get a failing grade.
A national survey, involving 1,547 adult Americans who were given a 20-question quiz about AI’s capabilities, found that only 16% of the test takers scored a passing grade of better than 60% on the quiz.
“The majority of Americans are AI illiterate,” Nicole DeCario and Oren Etzioni report today in a posting to PNW.ai, an information service provided by the institute, also known as AI2. Etzioni is AI2’s CEO, while DeCario leads AI2’s special projects team.
The researchers acknowledge that the extent of AI illiteracy shouldn’t be surprising. “AI is not part of our schools’ curricula, and the main source of information about it today, according to our survey, is YouTube and social media,” they write.
However, they argue that a basic understanding of how AI works is “critical for informing everyday decisions, adopting appropriate economic policies and maintaining national security.”