OpenAI’s new model “o1” has advanced reasoning, is based on a new paradigm

OpenAI is hunting for new capital - rumored to be $6.5 billion at a $150 billion valuation - so you know what that means: hype up the market by releasing a new model out of their toybox. Last Thursday OpenAI released “o1”, an AI model that is trained in advanced reasoning, making it better at logic puzzles, advanced math, and computer programming. Initial reviews by people way smarter than moi say it may be a new and promising direction for future AI models. First, it is an attempt to carve out a reasoning “core” from the massive models trained on vast amounts of data that are de rigeur today. Maybe AI systems of the future will be a collection of smaller models, each optimized for a particular type of intellectual task. Second, it gets it smarts not from ingesting a lot more data, but from thinking harder and more systematically at run time. o1 has been designed to think through problems step by step, and to check its work. Today’s models tend to pattern-match on the fly, and often get things wrong by what in a human we would call sloppiness.

The most succinct summation of o1’s abilities come from renowned UCLA professor of mathematics Terence Tao, known as the “Mozart of Math”: “[it is] roughly on par with… a mediocre but not completely incompetent graduate student.” Since Professor Tao attracts graduate students that are some of the best mathematical minds of their generation, this seems like praising with faint damns - or in other words, o1 is probably very, very good at advanced math.

OpenAI’s new model thinks harder to give smarter answers

Google’s Gemini Live voice assistant is free on all Android phones

Apple’s ho-hum hardware upgrades to the iPhone announced 2 weeks ago, along with its slooow rollout of the promised AI features, have given Google a chance to win market share in the smartphone market, and Google is pouncing on the opportunity. Last Thursday Google announced that its Gemini Live voice assistant (a smarter, perkier Siri) would be available for free on all Android phones. Since there is a growing consensus that voice interaction is the future of AI on smartphones, this puts pressure on Apple to deliver on its AI promises sooner rather than later. If you have an Android phone and want to replace Google’s lame Assistant with Gemini Live, just go to the app store and download and install the Gemini Live app.

Google’s Gemini Live voice assistant is free on all Android phones

Microsoft announces Copilot Wave 2 upgrade

Microsoft has been working hard on integrating partner OpenAI’s artificial intelligence models into its venerable Microsoft 365 suite of productivity apps (Word, Excel, etc.), branded as Copilot. Now they are releasing a second wave of AI upgrades, called (oh so cleverly) Copilot Wave 2. This will include advanced data analysis in Excel, better design in PowerPoint, inbox management in Outlook, and more. It also means that the user can design and deploy “agents” to automate common tasks, and generate powerful Python computer code just by asking for it in English.

Microsoft announces Copilot Wave 2 upgrades to productivity

Waymo partners with Uber to expand to more cities

Last week we reported that Waymo’s driverless taxis are 4 times safer than human drivers, as reflected in crash reports. This week Waymo is in the news announcing a possibly pivotal partnership with rideshare giant Uber. Waymo has agreed to use Uber’s technology for ride hailing in its next 2 expansion cities, Atlanta and Austin. This partnership will likely help Waymo roll out to cities faster, and Uber will be in excellent position to gradually phase out human drivers in favor of robots who are unlikely to strike or unionize.

Waymo partners with Uber to expand to more cities

Highlights of Oprah’s special on AI

Last Thursday, Oprah moderated a ballyhooed special on “AI and the Future of Us.” The self-made billionaire Queen of Talk was able to interview tech giants Bill Gates and OpenAI CEO Sam Altman, as well as current FBI Director Christopher Wray. The dominant tone was skepticism and wariness. She asked Altman if he was the most powerful and dangerous man in the world. Sam demurred. Oprah also dived deep on deepfakes (she’s against them, no surprise.) She brought on FBI Director Wray who warned against deepfakes, but also warned against AI-generated disinformation in the current US presidential election. Bill Gates tried to strike a more optimistic tone by pointing to the possible benefits that AI could bring in the fields of education and medicine. In prepared remarks, Oprah stated that the “AI genie is out of the bottle” and we must find ways to adapt to it.

Oprah expresses skepticism and fatalism about AI on her special

Google backs AI network to spot new wildfires from space

Google is backing a private consortium that will launch a series of satellites into space, carrying new sensors that will be analyzed by AI to spot new wildfires from space. Current satellite imaging technology can detect and track large, established wildfires easily, but is not designed to catch the small, new wildfires which if detected, could be more easily extinguished before they become larger and more destructive. The planned FireSat network will be able to detect wildfires as small as 16’x16’, the size of a school classroom.

Google backs AI powered satellite network to spot new fires from space

AI is paving the way to smoother streets using autonomous robots

The University of Liverpool School of Engineering has developed AI-powered autonomous robots for road repair, which have now spawned a spinoff company named Robotiz3d to commercialize the technology. Robotiz3d has been funded by a variety of public transportation agencies in the UK and Europe. The autonomous robot repair vehicle detects road cracks via AI analysis of camera data, then repairs the crack on the spot to prevent progression to a pothole, which is much more disruptive to traffic and more costly to repair.

AI is paving the way to smoother streets using autonomous robots

PaperQA2 conducts scientific literature review better than PhDs

Researchers recently announced Paper QA2, an opensource AI agent that conducts scientific literature searches on its own, at a level equal to or superior to PhDs and Post-docs in biology, as measured by both objective benchmarks and human raters. Authors of the paper predict that autonomous research agents will transform the way humans interact with the scientific literature.

PaperQA2 conducts scientific literature review better than PhDs

Oscar Health test drives OpenAI’s new o1 model, gives kudos

Oscar Health is a tech-forward health plan operating in 15 states. They had previously been using OpenAI’s GPT 4o for various health plan tasks, so when the o1 model was released last week, Oscar immediately began putting the new model through its paces. To their surprise, o1 was significantly better than the older model on claim pricing, application of clinical guidelines, detecting fraud/waste/abuse, clinical data extraction from the medical record, and automating administrative tasks.

Oscar Health test drives OpenAI’s new o1 model, and deems it a winner

AI tool reduces risk of death for hospitalized patients

Researchers in Canada have developed CHARTWatch, an AI tool that monitors the clinical status of high risk hospitalized patients and sends real-time alerts to clinicians, twice daily email updates to the nursing teams, and daily emails to the palliative care team. The results were a nearly 25% reduction in nonpalliative deaths (1.6% in the intervention period, versus 2.1% during the baseline period.)

AI tool reduces risk of death for hospitalized patients

AI can analyze patient EKGs to assess cardiac risk from chemo

Yale researchers have developed an AI system that can analyze EKGs of prospective chemotherapy patients and determine the risk of treatment-associated cardiac dysfunction. This system can be used to more accurately assess the risk of negative cardiac outcomes for patients considering cardiotoxic chemotherapy regimens.

AI can analyze EKGs to assess cardiac risk from chemo

Medscape offers free AI Scribe services

AI Scribe services are taking healthcare by storm, bidding fair to become the standard of care within a very few years. By using AI to transcribe doctor-patient interactions in real time, these services free physicians from the administrative burden of charting visit notes, a task which has been long identified as one of the chief factors driving rampant physician burnout. Fierce competition has driven down the monthly per-physician charges for these services, from $600/month a year ago, to $300/month 6 months ago, to as low as $69/month from some startups today. Now online health information service Medscape is lowering the bar even further, offering free AI Scribe services to members.

Medscape offers free AI Scribe services