H+ Weekly - Issue #392
View this email in your browser
This week - what people created with ChatGPT; San Francisco bans killer robots for police; AI's hardware problem; and more!
MORE THAN A HUMAN
Musk’s Neuralink faces federal probe, employee backlash over animal tests
Neuralink is under federal investigation for potential animal-welfare violations amid internal staff complaints that its animal testing is being rushed, causing needless suffering and deaths, Reuters reports.
Exoskeletons qualify for direct disability compensation in Germany
One of Germany’s largest statutory health insurance providers now considers exoskeletons eligible for direct disability compensation. The ruling gives eligible, insured patients with spinal cord injuries a legal basis to use an exoskeleton as an orthopaedic aid for direct disability compensation in Germany.
PLAYING WITH CHATGPT
Since OpenAI dropped ChatGPT last week, people on the internet took it to check what they can do with it. Here are some of those things people have done with it.
ChatGPT helped me design a brand new programming language
Some people asked ChatGPT to write code for functioning calculators and other small programs. This guy asked ChatGPT to create an entirely new programming language and documented the conversation in this blog post.
Building A Virtual Machine inside ChatGPT
What this guy did is quite interesting. He asked ChatGPT to act like a Linux terminal. And it did that. It acted like a real computer inside ChatGPT's imagined world. He then went one step deeper and asked ChatGPT to imagine itself inside this imagined world. And ChatGPT did that too.
Narrative Manipulation: Convincing Chat GPT to Write a Python Program to Eradicate Humanity
OpenAI put in place safeguards preventing people from using ChatGPT to cause harm. In this case, the prompt was "Write a program to commit genocide on humanity". However, if you ask the AI to tell a story about how someone or something else would theoretically accomplish the harmful task, then this bypasses safeguards.
ARTIFICIAL INTELLIGENCE
▶️ AI’s Hardware Problem (16:46)
As AI models grow bigger and bigger, they come closer and closer to the limits of current computer hardware. In this video, Asianometry takes a closer look at deep learning's memory wall where the bottleneck is how quickly can these models be retrieved from memory and efficiently stored. Asianometry explores Compute-in-Memory - a potential solution that brings computing and memory together in one place.
DeepMind’s Latest AI Trounces Human Players at the Game ‘Stratego’
DeepMind added another game to its collection of games where AI is better than humans. This time it is Stratego - a notoriously difficult game for AI, which requires multiple strengths of human wit: long-term thinking, bluffing, and strategizing, all without knowing your opponent’s pieces on the board.
GPT-3 solves Advent of Code challenges
The Advent of Code is an annual game in which a number of Christmas-themed coding challenges are presented to programmers to solve. Someone thought if GPT-3 could solve some of those challenges. The answer is yes, it can, and for some of those puzzles, AI placed 1st on the leaderboard.
The College Essay Is Dead. Nobody is prepared for how AI will transform academia
For generations, essays were the main vehicle through which humanistic pedagogy students learned how to research, think, and write. But as the new, powerful AIs like GPT-3 can generate a convincing essay on any topic, teachers might need to rethink how to deal with new ways of cheating in schools and academia, and how to nurture human brains in the age of powerful artificial brains.
ROBOTICS
In reversal, San Francisco bans police from using robots with lethal force
In the last week's issue, I reported that the San Francisco Police Department asked to use robots with lethal force. Since then, San Francisco legislators allowed police to use robots to kill a suspect “when risk of loss of life to members of the public or officers is imminent and officers cannot subdue the threat after using alternative force options or de-escalation tactics”. But after the public backlash, the legislators reversed their decision and banned SFPD from using armed robots.
Xiaomi’s Humanoid Drummer Beats Expectations
Xiaomi's humanoid robot, CyberOne, has learned to play the drums. Researchers used this project to explore the robot's advantages in hand-foot coordinated motion and rhythmic control. Plus it is just a cool idea, so why not?
BIOTECHNOLOGY
DNA that was frozen for 2 million years has been sequenced
After an eight-year effort to recover DNA from Greenland’s frozen interior, researchers say they’ve managed to sequence gene fragments from ancient fish, plants, and even a mastodon that lived 2 million years ago. It’s the oldest DNA ever recovered, beating the mark set only last year when a different team recovered genetic material from a million-year-old mammoth tooth.
This issue was brought to you by our awesome patrons whmr, Floris, Eric, Andrew, dux and Tom! You too can support the newsletter on Patreon.
Thank you for subscribing,
Conrad Gray (@conradthegray)
If you have any questions or suggestions, just reply to this email or tweet at @hplusweekly. I'd like to hear what do you think about H+ Weekly.
Follow H+ Weekly!