AI passes the US Medical Licence Exam

Hey there,

Happy new year! Hope you had a recharging holiday break ☺️. I definitely had a good time. Here’s a gif from Christmas in Rome with our family dog Martin, and NYE in the windy middle of absolute nowhere, Denmark.

I’m trying something new for the first 2023 edition of Tech Pizza! My data analysis showed that Tech Pizza has become constantly longer and longer during 2022, so I want to try making one edition without my usual super long speech about a specific piece of news.

I gave a few lines of space to all the important stuff that happened in the past few weeks and divided it into different topics. For some of the news I had to resist the urge to write 35078 words, but maybe I’ll do that in a separate email.

Reply to this email and let me know your thoughts (or just with me happy new year!).

Let’s get started 🕺

🤖 AI 🤖

🤖 🧑‍⚕️ Google’s Deepmind fine-tuned an AI model similar to ChatGPT to answer medical questions. It’s called FlanPaLM, and it has 67.6% accuracy on the US Medical Licence Exam questions. To pass, it’s enough to have 60% accuracy. Yes, this AI could pass the test to be a licensed US doctor.

How does it compare to doctors? They had a competition: 🤖 VS🧑‍⚕️. Both groups were asked 140 medical questions. Doctors agreed 92.9% of the time with the medical consensus. Deepmind’s AI scored 92.6%. That’s a veeeeery thin margin.

Lay users were asked to rate the usefulness of the answers. They rated 91.1% of the doctors' answers as "Helpful", and only 80.3% of the AI's. There is still progress to be made, but the direction is clear.

Funny anecdote: I found this paper 10 minutes after replying to a LinkedIn comment saying that doctors were definitely not going to be replaced by AI. The future is accelerating.

🚨 Google’s management declared ChatGPT a “code red.” This means that they see OpenAI’s technology as an existential threat to Google, and are now pulling resources from different parts of the company to fight back. I said multiple times that I’m 100% sure Google has even better technology than OpenAI (some of it it’s been described too) but has a fundamental business problem: how do they use it without killing their current “click on this link” business model? How the hell do you show ads in an AI response?

🤖 🔎 Microsoft is looking into integrating OpenAI’s chatGPT into Bing, its search engine. It means total sense, as Microsoft is pouring billions to fund OpenAI’s research and Bing never really competed with Google (also, see news above).

🤖 🎙️ Apple launches an AI tool for authors to generate an audiobook from…a normal book. From the demos, it seems the most natural-sounding AI voice I’ve ever heard. Also, in true Apple fashion, they perform a final manual inspection of the audiobook to ensure it meets their quality standards. Cool.

🤖 🌿 Google’s Deepmind used AI to cut the energy expense of commercial buildings’ cooling systems by between 9 and 13%. They detailed their approach in a scientific paper. This is the follow-up of their work on cutting Data Centers’ consumption, applied to a broader case. It’s quite important since buildings accounted for 37% of energy and process-related CO2 emissions in 2021.

🔐 CRYPTO 🔐

🤑 According to a new research by Dune Analytics, 58% of NFT Trading Volume on Ethereum Is Wash Trading: a form of market manipulation where the buyer and seller in a transaction are the same or collude together. That’s illegal in regulated financial markets, by the way.

🟧 Donald Trump dropped an NFT collection. I was unsure if keeping this news for the “crypto” or “fun” section of the newsletter.

🏦 BIG TECH 🏦

🕵️ ByteDance (TikTok’s parent company) confirmed that they used their app to monitor the location of US journalists investigating the company. This was part of an initiative to unearth the source of leaks inside the company. ByteDance fired Chris Lepitak, its chief internal auditor who led the team responsible for them.

🚫 🚗 A new California law prevents car manufacturers from "deceptively naming or marketing" a car as self-driving if it's only equipped with partial automation features that still require human drivers to pay attention and handle driving. They’re looking at you, Elon (we talked about Tesla’s marketing lies right before Christmas).

💸 Meta has agreed to pay $725 million to resolve a class-action lawsuit for breaching customers’ privacy in regard to the Cambridge Analytica scandal.

✍️ Instagram launches Notes, its first text-only feature. Are they trying to compete with Twitter given its total state of chaos?

😋 Tiramisù 😋

The sweet part of the newsletter: fun news from the crazy world of tech.

♥️ 🤖 This guy used ChatGPT to write a love poem for his Tinder match. Told ya the applications of AI are endless.

🟧 LOVED this video about Trump’s NFTs 😆

💰 Not really funny, but very interesting visualization of why the ultra-rich are inevitable.

🔮 This image pretty much summarizes how 2023 will be:

You reached the end of this edition 😢

I’d love to know if you liked it. Let me know by replying to this email!

I’ll talk to you next week.

Ciao 👋

What our subscribers say

Agatha Peptea

I totally recommend it, great info and also great fun! As you can deduct from above, I declare myself a fan🙂 happy to read this condensed newsletter about tech!
Cultural Change Director - E.ON

Lubov Ostrovskaya

I feel more empowered intellectually 🙂
Business development - FNA SPA

Subscribe to our newsletter

Your 10 min shortcut to understand how tech impacts society and business, available in audio, video or text format

Check Icon - Newsletter X Webflow Template
Don’t worry, One highly curated email per week, no SPAM
Cookies Preferences
Close Cookie Preference Manager
Cookie Settings
By clicking “Accept All Cookies”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage and assist in our marketing efforts. More info
Strictly Necessary (Always Active)
Cookies required to enable basic website functionality.
Made by Flinch 77
Oops! Something went wrong while submitting the form.