Toggle light / dark theme

Connor Leahy from Conjecture joins the podcast to discuss AI progress, chimps, memes, and markets. Learn more about Connor’s work at https://conjecture.dev.

Timestamps:
00:00 Introduction.
01:00 Defining artificial general intelligence.
04:52 What makes humans more powerful than chimps?
17:23 Would AIs have to be social to be intelligent?
20:29 Importing humanity’s memes into AIs.
23:07 How do we measure progress in AI?
42:39 Gut feelings about AI progress.
47:29 Connor’s predictions about AGI
52:44 Is predicting AGI soon betting against the market?
57:43 How accurate are prediction markets about AGI?

Summary: Combining neuroimaging and EEG data, researchers recorded the neural activity of people while listening to a piece of music. Using machine learning technology, the data was translated to reconstruct and identify the specific piece of music the test subjects were listening to.

Source: University of Essex.

A new technique for monitoring brain waves can identify the music someone is listening to.

Summary: A newly developed machine learning model can predict the words a person is about to speak based on their neural activity recorded by a minimally invasive neuroprosthetic device.

Source: HSE

Researchers from HSE University and the Moscow State University of Medicine and Dentistry have developed a machine learning model that can predict the word about to be uttered by a subject based on their neural activity recorded with a small set of minimally invasive electrodes.

Boston Dynamics’ Atlas—the world’s most advanced humanoid robot—is learning some new tricks. The company has finally given Atlas some proper hands, and in Boston Dynamics’ latest YouTube video, Atlas is attempting to do some actual work. It also released another behind-the-scenes video showing some of the work that goes into Atlas. And when things don’t go right, we see some spectacular slams the robot takes in its efforts to advance humanoid robotics.

As a humanoid robot, Atlas has mostly been focused on locomotion, starting with walking in a lab, then walking on every kind of unstable terrain imaginable, then doing some sick parkour tricks. Locomotion is all about the legs, though, and the upper half seemed mostly like an afterthought, with the arms only used to swing around for balance. Atlas previously didn’t even have hands— the last time we saw it, there were only two incomplete-looking ball grippers at the end of its arms.

This newest iteration of the robot has actual grippers. They’re simple clamp-style hands with a wrist and a single moving finger, but that’s good enough for picking things up. The goal of this video is moving “inertially significant” objects—not just picking up light boxes, but objects that are so heavy they can throw Atlas off-balance. This includes things like a big plank, a bag full of tools, and a barbell with two 10-pound weights. Atlas is learning all about those “equal and opposite forces” in the world.

With a stylized celebration to celebrate at the end.

Boston Dynamics has done it once again. After demonstrating the extreme capabilities of its bipedal robot, Atlas, flawlessly executing parkour tricks, the company has now released a video where you will fall in love with the robot for doing what one hates the most–climbing down from a high platform or ladder to get the tool you need.

Needless to say, the video is shot inside Boston Dynamics’ controlled facility and results from hours of perspiration and many broken robotic appendages, something we have covered before.

In this vignette, a human is shown working on a high platform and realizes that he has forgotten his tool bag, which happens very often. In a world where a robot like Atlas is indeed at our beck and call, one can ask him to hand over the toolbag, which he does effortlessly.

In the world of spreadsheets and data analysis, a new player has emerged to shake up the game. Akkio, the easy-to-use AI company, has launched Chat Data Prep, a revolutionary machine learning platform that allows users to transform data using ordinary conversational language.

Gone are the days of struggling with complicated formulas and formatting commands in Excel. With Akkio’s Chat Data Prep, users can simply type in conversational language to make changes to their spreadsheet data. Leveraging AI and large language models, the platform interprets the user’s requests and makes the necessary changes to the data.

According to Jonathon Reilly, co-founder of Akkio, this new method of interacting with data results in a 10-fold reduction in the time it takes to prepare data for analysis. With Chat Data Prep, users can reformat dates, perform time-based math operations, and even fix messy data fields with a simple conversational command.

In a recent interview, Altman discussed hype surrounding the as yet unannounced GPT-4 but refused to confirm if the model will even be released this year.

OpenAI CEO Sam Altman has addressed rumors regarding GPT-4 — the company’s as yet unreleased language model and latest in the GPT-series that forms the foundation of AI chatbot ChatGPT — saying that “people are begging to be disappointed and they will be.”

During an interview with StrictlyVC, Altman was asked if GPT-4 will come out in the first quarter or half of the year, as many expect. He responded by offering no certain timeframe. “It’ll come out at some point, when we are confident we can do it safely and responsibly,” he said.


The success that ChatGPT has had, at least in generating public interest, has had the inevitable consequence of prompting some writers to question its credentials and generally pour tepid if not actually cold water over what it can do. The latest of these is Will Knight writing in the January 13, 2023 edition of Wired. “ChatGPT Has Investors Drooling – but Can It Bring Home the Bacon?”.

In that article he makes two observations that merit closer attention, one of which I think has merit and the other of which I think harks back to a Dreyfus-like What Computers Still Can’t Do mentality. And both can be seen as examples of Schadenfreude.

Right at the end of the article Wright makes a legitimate point that he has gleaned from Phil Libin who was the CEO of the note-taking app Evernote from 2007–2015. Wright, summarising some of the downsides Libin anticipates, says One is that ChatGPT and other generative AI models are currently created by scraping content made by humans from the web, but are increasingly contributing to the text and images found online. All of these models are about to shit all over their own training data, he [Libin] says. ‘We’re about to be flooded with a tsunami of bullshit.’