Top 150 AI tools in 2023

AI is rapidly going mainstream. AI sites and tools generate over 3 billion monthly visits. Just over half of that (1.67 billion) is for OpenAI, making ChatGPT more popular than Netflix (1.5 billion monthly visitors.) (!?!) Top uses of AI tools are for Writing and Editing, Image Generation, Education (mostly homework help), Characters (socializing with AI “friends” - and “girlfriends”), and Research (from academic papers to settling bar bets.) Click the link for some intriguing charts to info-snack on.

AI is going mainstream

Apple’s pricey VisionPro “mixed reality” headset is here to stay

Apple’s much-touted VisionPro “mixed reality” (what The Elders used to call “augmented reality”, i.e. you can see both the real world and a virtual overlay) headset is finally on sale, and while it is not perfect, it is likely plenty good enough to set a new standard in the AR category. Click the links for some initial examples of the VisionPro in use. Want one? It will set you back a cool 3500 bucks.

21 ways to use your new VisionPro

A Tech reporter’s first weekend with the VisionPro

Microsoft is partnering to bring AI to the newsroom

Microsoft wants to be the infrastructure behind tomorrow’s AI-powered newsroom. Microsoft has inked collaboration agreements with Craig Newmark School of Journalism at CUNY, the GroundTruth Project for minority voices in the news, the Online News Association, and has made a pledge of major support to startup online news site Semafor. Microsoft will provide access to AI tools for news collection, while human reporters will write the stories. Until… maybe, just maybe, until the bots are better…ya think?

Microsoft partners with news orgs to bring AI to the newsroom

Deepfake scammer grabs $25 million in AI heist

Deepfaking Taylor Swift porn is so last week. The new kings of Deepfake are the criminals who tricked a Hong Kong employee into transferring $25 million of company funds to fraudulent bank accounts. The employee was invited onto a video conference with what appeared to be the company’s CFO and several other firm employees. All of the participants in the video call except the scammed employee were AI digital clones, whose appearance and voices were copied from publicly available photos and videos of the real company executives. This heist is the first of its kind - it won’t be the last.

Deepfake scammer grabs $25 million in AI heist

Ancient scrolls buried by Vesuvius deciphered by AI

When Mount Vesuvius exploded in 79 AD, it buried the wealthy towns of Pompeii and Herculaneum in ash - including thousands of papyrus scrolls which were singed to a crisp as well. For the 500 years since the towns’ ruins were rediscovered, the scrolls were unreadable, due to their hyper-fragile charred condition. A few years ago, the scrolls were CT-scanned, and work began to use the scans to decipher the texts. Last year, tech entrepreneur Nat Friedman funded a million-dollar Vesuvius Challenge contest, to engage enthusiasts around the world in the quest to devise a method for reading the texts. On Monday, Friedman announced that a trio of 3 college students (!?!) had created a neural network that successfully read part of one scroll, and so were awarded the $700,000 Grand Prize. And what did you do over Winter Break, hmmm???

Twitter/X: 3 college students decipher 2000-year-old mystery scrolls

Ars Tecnica article on the backstory

Roblox announces in-game text translation for 16 languages

Online game platform Roblox announces instantaneous translation of in-game texts between players for 16 languages. This opens the door to more engaging international teams and matches. Next up, says the company, is instantaneous voice translation, that uses your own voice and tone in the new language.

Roblox announces in-game instant text translation for 16 languages

Fake Biden robocall companies targeted by NH AG and the FCC

It takes a special kind of stupid to draw attention to yourself while breaking the law. Fortunately for US democracy, the Texas companies behind the notorious fake Joe Biden robocalls in the New Hampshire primary election had that special quality in abundance. Texas-based Life Corporation has been served a cease-and-desist letter from the NH Attorney General, and more trouble is sure to come. The FCC has also served the company with a cease-and-desist order, and has served the other company involved, Lingo Telecom, with a similar order. The FCC is vowing to make almost all voice cloned robocalls illegal.

Fake Biden robocall companies in a heap of trouble with FCC and NH AG

Kaiser study finds AI scribes are good, and need to get better

Currently, physicians are slaves to their EMR. Multiple studies have shown that “EMR fatigue” is a major factor in physician burnout. And the constant key-clicking during office visits is a major annoyance to patients as well. Recently, a number of AI solutions known as AI scribes have hit the market, in which a real-time transcript of the physician-patient interaction is immediately summarized by AI into a draft visit note. Now Kaiser has analyzed 300,000 encounters transcribed by AI, and find that 1) the quality of the draft notes were good, but that the notes still frequently needed editing; 2) physicians were favorable, finding that the AI scribe freed them up to interact with the patient, and reduced after-hours documentation tasks; and 3) patients were favorable, liking the extra attention from the physician in the visit. Expect this to be a hotly competitive and rapidly developing sector of the AI market.

Kaiser study concludes AI scribes are good, and need to get better

Mass General uses AI to optimize palliative care utilization

Boston-based MGB used a commercially available AI tool to identify patients that were eligible for hospice care. Transition to hospice care gives patients a wealth of additional services above standard Medicare benefits, and has major positive impact on health plan finances as well. In a 6 month pilot at a single hospital, the AI tool identified 17 additional hospice-eligible patients compared with traditional case finding methods. This represented an estimated $2 million annual savings to the health plan, plus improved quality of life for the identified patients.

MGB uses AI to optimize palliative care utilization

Stanford team builds SOTA medical question answerer with RAG

A team at Stanford has built Almanac, an AI Large Language Model that outperforms existing AI models such as ChatGPT in answering complex medical questions. The secret? Almanac actually searches a pre-processed medical data base for the information needed to answer the question. Sounds simple. It’s not. Retrieval Augmented Generation (RAG) (aka looking up the answer) is at the cutting edge of AI research. It’s State of the Art (SOTA). The problem is, LLMs hallucinate (tell fibs.) A lot. Even when they look up the answer, they might lie about what they found. Almanac doesn’t, at least not as much as the others. Progress. 

Stanford team builds SOTA medical question answerer with RAG

CRISPR cures hereditary angiodema. Just another Tuesday.

CRISPR, the Nobel Prize-winning gene splicing tool, seems to be racking up cures for genetic diseases almost every week these days. It seems like just yesterday that the FDA approved a CRISPR cure for sickle cell disease. (Nope. It was a full 2 months ago.) This week, we have a paper in the New England Journal of Medicine about a CRISPR cure for hereditary angioedema, a genetic disorder that can be fatal if not properly and continuously treated, especially in childhood. Expect the torrid pace of CRISPR cures to continue for a while, cuz lots of teams are working on lots of hereditary diseases.

CRISPR cures yet another genetic disease.