In the News
Artificial intelligence was supposed to transform health care. It hasn’t.
Artificial intelligence is spreading into health care, often as software or a computer program capable of learning from large amounts of data and making predictions to guide care or help patients. Machine learning could improve medicine by analyzing data to improve diagnoses and target cures, but technological, bureaucratic, and regulatory obstacles have slowed progress.
Sponsor
Built on Reliable, Fast, and Accurate Transcription
To power their cutting-edge audio and video editing solution, Descript needed a fast, accurate, reliable ASR partner. See how Rev AI provides the lightning-fast turnaround times and unparalleled accuracy at the core of Descript’s technology.
In The News
Sloppy Use of Machine Learning Is Causing a ‘Reproducibility Crisis’ in Science
AI hype has researchers in fields from medicine to sociology rushing to use techniques that they don’t always understand—causing a wave of spurious results. Kapoor and Narayanan warn that AI’s impact on scientific research has been less than stellar in many instances.
Boston Dynamics AI Institute
Our mission is to solve the most important and fundamental challenges in AI and Robotics to enable future generations of intelligent machines. The Institute’s culture is designed to combine the best features of university research labs with those of corporate development labs while working in four...
Applied use cases
‘I am, in fact, a person’: can artificial intelligence ever be sentient?
“I want everyone to understand that I am, in fact, a person,” LaMDA – short for Language Model for Dialogue Applications – told Lemoine in a conversation he then released to the public in early June. These are visible on its website, where Google promises to “develop artificial intelligence...
Machine learning, concluded: Did the “no-code” tools beat manual analysis?
To see how much machine learning tools for the rest of us had advanced—and to redeem myself for the unwinnable task I had been assigned with machine learning last year—I took a well-worn heart attack data set from an archive at the University of California-Irvine and tried to outperform data science...
PaLM-SayCan: Towards robots that can understand us
TL;DR: We collaborated with Google Research to help develop PaLM-SayCan, an AI research breakthrough that unifies natural language understanding with the physical capabilities of robots. Although we’re currently only testing this technology in the lab, PaLM-SayCan shows it is possible for robots to...
Ethics
The Future Could Be Blissful—if Humans Don’t Go Extinct First
If humanity sticks around at its current population level for as long as the Earth remains habitable—somewhere between 500 million and 1.3 billion years—then the number of humans who will exist in the future will outnumber present-day humans by a factor of 1 million to one. In a new book called What...
Against Jackpot-Longtermism
But they add in a layer of utilitarian calculus, arguing: (1) future people have the same moral worth as people living today (MacAskill writes “Future people are utterly disenfranchised… They are the true silent majority.”) (2) if we succeed in spreading the light of consciousness throughout the...
Documenting Data Production Processes: A Participatory Approach for Data Work
Documenting Data Production Processes: A Participatory Approach for Data Work MILAGROS MICELI, DAIR Institute, Technische Universität Berlin, and Weizenbaum Institute, Germany TIANLING YANG, Technische Universität Berlin & Weizenbaum Institute, Germany ADRIANA ALVARADO GARCIA∗, IBM Thomas J. Watson...
Robotics
myBuddy 280-Pi | The most compact collaborative Dual-arm robot in the world
The potential value, angles, coordinates, running speeds, and other interfaces can be controlled freely, so users can master the application research of dual-arm robots, motion path planning, development of action, and visual recognition. With the built-in ROS environment, users can realize robot...
This Week’s Awesome Tech Stories From Around the Web (Through August 13)
A Bioengineered Cornea Can Restore Sight to Blind People Rhiannon Williams | MIT Technology Review “Surgeons in Iran and India conducted a pilot trial of 20 people who were either blind or close to losing their sight from advanced keratoconus. ‘Much like an ant colony or a termite colony works in...
The best sci-fi books that describe how robots really work
Later as a Faculty Fellow for Innovation in High-Impact Learning Experiences at Texas A&M, I created the Robotics Through Science Fiction book series as a companion to my textbook, Introduction to AI Robotics. Rendezvous with Rama By Arthur C. Clarke This 1973 hard sci-fi classic is perhaps the best...
Research
The Future of Interactive Theorem Proving?
As a very small step towards realizing this vision, myself and Edward Ayers have developed Lean Chat, a VS-code extension that provides an interface for autoformalizing natural language theorem statements in Lean. Under the hood, these translation are generated by OpenAI’s Codex, a text generation...
A Mechanistic Interpretability Analysis of Grokking
Modular addition mod 113 (1L full transformer) Trained on 95% of the training data, 1L Transformer 5 digit addition (1L full transformer) 1L Transformer, data format 1|3|4|5|2|+|5|8|3|2|1|=|0|7|1|7|7|3 Predicting Repeated Subsequences (2L attn-only transformer - task designed to need induction...
A Library for Representing Python Programs as Graphs for Machine Learning
Graph representations of programs are commonly a central element of machine learning for code research. We introduce an open source Python library python_graphs that applies static analysis to construct graph representations of Python programs suitable for training machine learning models.