Google Genesis: The Empire Strikes Back!

Google was King of AI until ChatGPT stole its thunder. When OpenAI’s chatbot burst on the scene last November, all the companies with big AI R&D programs (Google, Meta/Facebook, Amazon, Apple) were caught flat-footed. Now Google is aiming to retake the crown, releasing its new multimodal LLM, codenamed Genesis, to a select group of companies that are serving as alpha users. Google’s Deepmind CEO is talking trash, saying Genesis will be wa-a-ay better than ChatGPT, able to handle not just text, but audio, images, video - the Full Monte. Was ChatGPT just a flash in the pan???

Google nears release of next-generation LLM, Gemini

Meanwhile, Young Skywalker races to outflank the Death Star!

Oh, Hell to the NO, says young Luke - er, Sam Altman! Credible sources say that OpenAI is putting the pedal to the metal to try to release a new multimodal upgrade to ChatGPT, codenamed Gobi, ahead of Google’s Gemini. This model has been alpha-tested by a single company, Be My Eyes, that aims to assist the vision-impaired with their daily activities. (Google rasps, “Luuuke -I am your fatherrrr…”)

OpenAI races to launch multimodal upgrade before Google

Elite consulting group tests GPT-4 and the results are Wow!

A consortium of researchers from Harvard, MIT, and Wharton business schools ran a formal study of the impact of Generative AI on the productivity and quality of output for knowledge workers. Top consultancy Boston Consulting Group worked with the research team to enroll 758 of their consultants to gauge the effect of AI on 18 realistic, complex, knowledge-intensive tasks similar to their daily work.

The results were astonishing. Consultants who used AI completed 12% more tasks, 25% faster, with 40% better quality scores than their personal non-AI baseline, and as compared to consultants who did not use AI.

AI had a leveling up effect, improving the performance of lower performers more than it helped higher performers.

Working styles varied from “centaurs” who used AI for certain tasks and completed other tasks with traditional methods, to “cyborgs” who fluidly integrated more or less AI into each task they tackled, depending on the situation.

These results, if they hold up in the real world, have huge implications for the future of knowledge work. So - are you a centaur or a cyborg?

Study shows AI turbocharges knowledge work

Singapore arms firefighters with AR glasses

Singapore has embarked on a 2-year initiative to equip its firefighters with 5G-enabled smart glasses, to deliver AI directly to the firefighter in the field. The first project will allow the firefighter to automatically check their equipment for deficiencies and defects, just by looking at it.

Future projects are expected to use embedded voice-activated technology and heads-up display to make the smart glasses a mobile dashboard and mapping device for firefighters at a blaze.

Singapore arms firefighters with AR glasses

Character AI is catching up to ChatGPT

Premier VC fund a16z flexes its analytic prowess again (see their article in the AI in Medicine section) to rank the current crop of AI apps by monthly usage. Unsurprisingly, ChatGPT leads to pack - by a lot. Less expected, the #2 slot, at about half ChatGPT’s volume, is held by “rent-a-friendbot” site, Character AI. And Character is moving up fast, helped by its mobile-friendly design. Character’s engagement numbers are insane - the average user spends 2 hours a day on the app (4 sessions of about a half-hour each day, talking to your robot friend.) Why am I getting strong “Her” (the movie) vibes?

Character.AI challenges ChatGPT for top AI app

Coke releases AI-generated soft drink and can design

Coke, who made one of the best early AI-generated TV ads (remember the museum paintings that came alive?) has now released a new futuristic zero-calorie drink that was created with help from an AI chatbot, who assisted in everything from choosing a flavor to designing the can.

Coke designs the future with AI

Teachers are going all in on Generative AI

Just a few short months ago, Generative AI was being painted as the doom of education - chatbots were “cheatbots” that would allow lazy students to graduate without learning anything. This fall, chatbots are swarming into the classroom - but they are being brought by the teachers. Teachers have quickly caught on that chatbots can automate the boring parts of the job, like creating lesson plans, or tweaking math word problems to pique student interest with pop-cultural references. A nifty idea - but are the teachers cheating???

(Only) the best humans outperform AI in creative thinking

In this paper from this week’s Nature, researchers compared the performance of 3 different chatbots to human participants in a standardized test of creative divergent thinking. The result was that the chatbots achieved higher average scores, but the highest individual scores were almost all held by humans. As were all the lowest scores. So, while the chatbots achieved a relatively consistent and moderately high level of performance, humans were all over the map, from geniuses to village idiots.

(Only) the best humans are more creative than chatbots

Google’s antitrust trial may rewrite our future

This NYT opinion piece by Columbia Law Professor Tim Wu, who has held posts at the Supreme Court and the White House, argues that the US Government’s current antitrust suit against Google is likely to redraw the rules of competition in the tech sector, to the advantage of smaller, newer, nimbler companies who might otherwise be crushed by the might of the current massive incumbents (collectively known as MAANG from their first initials.) He notes - ironies of ironies - that Google itself (founded in 1998) owes its survival to the aftermath of the antitrust suit against Microsoft that same year, which chastened Microsoft into less ruthless behavior against upstarts.

Google’s antitrust trial may rewrite the future of tech

VC firm’s white paper maps AI “Jobs To Be Done” in Healthcare

Marc Andreesen is a very smart guy - as he will be the first to tell you. After inventing the internet browser, he founded a premier VC company known as a16z. Now the team at a16z has released a white paper titled “Commercializing AI in Healthcare: The Jobs to be Done.” And gotta admit - it’s pretty good.

It identifies 19 areas of opportunity for AI in healthcare, mapped onto 2 axes - clinical vs. nonclinical, and patient-facing vs. professional facing. In some areas, like Medical Image Analysis, AI is already far advanced, and will soon be making major contributions to patient care. In other domains, like Scribing (automated visit notes) the technology is far ahead of the adoption curve. And in still other segments, like Clinical Decision Support, the technology lags far behind the promise. In the end, a major contribution of this white paper is just giving a roadmap of the automatable functions of the sprawling, complex beast that is US healthcare. (see graphic below)

What hath Marc wrought - AI “Jobs to be Done” in Healthcare

Summary on Twitter/X

Oracle announces an AI “clinical assistant” for its EMR

Database mega-vendor Oracle has gotten into apps lately, and it has its own electronic medical record (EMR). Now it’s adding AI features to its EMR, such as automated visit notes, or scribing. This echoes the blockbuster announcement last April that Microsoft and Epic would partner to put OpenAI’s industry-leading AI into Epic’s industry-leading EMR. Gonna be a whole lotta scribing going on some day soon…

Oracle announces voice-activated AI clinical assistant for its EMR

Microsoft and Epic partner to put AI into the EMR (April 2023)

Stanford researchers find AI summaries beat humans’

Concise, pointed summaries of a patient’s clinical information is the lifeblood of medical communication, and a major drain on clinician time. Now Stanford researchers have found that AI-generated clinical summaries can be more complete and more correct than summaries generated by human experts.

AI summarizes clinical text better than expert humans

AI retinal scanner detects eye disease and risk of Parkinson’s

In this week’s Nature, researchers from the UK, US, and Spain reported on an AI model that examines retinal images, and can then make diagnoses of eye diseases, cardiovascular diseases, and neurological conditions. The authors make the point that the retina allows us to directly visualize elements of the circulatory system and the brain, as well as components of the eye, and so can serve very literally as a window on the inner workings of the body.

AI model scans retinal images for systemic disease