OpenAI hits $1.3 billion annual revenues

Continuing its stratospheric growth, OpenAI is now making $130 million a month, compared to last year’s $28 million for the whole year. OpenAI rival Anthropic, developer of the other indie LLM, Claude, is only projected to make $100 million this year. It may be raining cash on AI companies these days, but OpenAI has the biggest bucket.

OpenAI revenues skyrocket to $1.3 billion

Air Street Capital releases its 6th annual “State of AI” report

Harking back to the legendary Mary Meeker’s industry-defining annual Internet Trends Report, the VC dudes at Air Street Capital have just released their sixth annual “State of AI Report.” It’s big, it’s meaty, it’s… ugly. Meeker told her story in stunning infographics, the dudebros… have a lot of text in bullet points. Duuudes - 1990 called and it wants its PowerPoint back. And yet, it’s still a must read, summarizing all the major trends and events of this past, very eventful year.

Air Street Capital’s 6th State of AI report

Adobe-led consortium creates symbol to tag AI content

In this new world of AI-generated deepfakes and voice clones, the issue of being able to identify AI-generated content is increasingly important. Adobe, developer of content tools like the PDF format and Photoshop, has a big stake in ensuring the integrity of online documents and images. Now, in consortium with Microsoft and other large online companies, Adobe has released a new symbol for tagging AI-generated content. So the user can beware. It’s ugly and confusing - what the heck does the “CR” even stand for??? - but hey, it’s a start.

Consortium releases tag for AI-generated content

Nvidia and Foxconn ally to build supercomputing “AI Factories”

Industry-leading GPU developer Nvidia is teaming up with Taiwanese electronics manufacturing giant Foxconn to build massive supercomputing facilities to accelerate development of self-driving cars. This comes on the heels of Tesla’s announcement of its own billion-dollar “Dojo” supercomputer. Safe to say, if these billionaires have their way, the car you own now will be the last one you drive yourself.

Nvidia and Foxconn team up on supercomputers for self-driving cars

AI spots its first supernova

Supernovas provide a wealth of information on how our universe works, but finding them is a tedious, manual process. Over the past 6 years, astronomers estimate that they have spent over 2,200 hours visually inspecting blobs of light to find the supernovae. Now, an AI called Bright Transient Survey Bot - or BTSBot for all you K-pop fans - has automated the process, and recently found its first “Look, Ma, no hands!” verified supernova. Now astronomers can skip the drudgery and go right to the data-collection and theorizing. Or spend more time on academic politics, so they can get tenured. Whichever.

AI spots its first supernova

Is ChatGPT better at therapy then your therapist?

Researchers at Carnegie Mellon and UC Santa Barbara studied whether, with proper prompting, an LLM could analyze cognitive distortions in patient speech. Turns out, GPT-4 is great at it, achieving perfect scores on analysis of 70-85% of patient samples, and only completely missing the boat on 5%. Given the dire shortage of psychotherapists, maybe a robot psychiatrist is the wave of the future.

GPT-4 detects cognitive distortions of psychotherapy patients

NYC Mayor faces backlash for using AI to make robocalls in languages he doesn’t speak

New Yorkers were surprised recently to receive automated calls from their mayor, not just in English, but in Spanish. And Yiddish. And in Mandarin Chinese. Impressed but skeptical, constituents pressed Hizzoner on his linguistic prowess, and he fessed up that he had used AI voice cloning. Outraged that someone in New York City - New York City! - might be less than completely aboveboard about something, they raised an outcry. Hizzoner, or someone who sounded very much like him, apologized.

NYC Mayor uses voice cloning

GPT-4 can guess your personal data from your public posts

Privacy is dead. Researchers show that GPT-4 can guess your private information (location, income, sex, race/ethnicity) just from the information you share in your public posts online, with an accuracy of 85%. And it can do it 100x faster and 240x more cheaply than humans.

GPT-4 can guess your personal data

AI diagnoses brain tumors on the operating table

There’s a reason we call any complex task “brain surgery.” Cutting out just the right part of the gray glop we call our brain is maddeningly difficult, but crucial to a successful outcome. In the past, neurosurgeons had to guess how much to cut out around a brain tumor to make sure they got it all. Now, scientists in the Netherlands have developed an AI that can analyze the tumor’s DNA while the patient is still on the operating table, to determine how aggressive the cancer is, and therefore how widely the surgeon should cut around the visible tumor so that no malignant cells stay behind. This real-time feedback may soon save lives - and brains - around the world.

AI analyzes brain cancers during the operation

Harvard and Oxford develop AI that predicts virus mutations

The pandemic taught us a lot. For one thing, those pesky, deadly viruses don’t stay still - they mutate and evolve. This can make vaccines ineffective, and public health becomes an arms race between the vaccine developers and the virus. But as the great Wayne Gretsky once advised, “Don’t skate to where the puck is, skate to where it’s going to be.” Harvard Medical School and Oxford University have decided to do just that, developing an AI that predicts how a virus will evolve, so that vaccine developers have a head start on producing the vaccine that will be needed next year, today.

AI makes vaccines “future-proof”