Messages
7/7/2020
10 : 46
Import AI 204: Chinese elevator surveillance; Enron emails train polite AI; and a pessimistic view on AGI
If we wanted to train a generative vision model of the solar system's major planets, how many machines might we deploy with what (plausible) imaging resolutions to collect the data? View this email
6/29/2020
11 : 7
Import AI 203: DeepMind releases a synthetic robot dog; first-person AI datasets; plus, harvesting kiwifruit with robots
If we continue to develop various AI*search techniques, then eventually we'll start fielding systems that learn their own multi-dimensional reward functions. How interpretable might these signals
6/22/2020
10 : 46
Import AI 202: Baidu leaves PAI; can ImageNet live forever with better labels?; and what a badly upscaled Obama photo tells us about data bias
How might AI combine with game engines to create smart, procedural gameworlds? And what might the intelligence vs graphics tradeoff become as capabilities advance? View this email in your browser
6/16/2020
9 : 46
Import AI 201: the facial recognition rebellion; how Amazon Go sees people; and the past&present of YOLO
Along with axis will future AI systems be 'dumb'? View this email in your browser Welcome to Import AI, a newsletter about artificial intelligence. Forward this email to give your chums an AI
6/8/2020
11 : 6
Import AI 200: Pedestrian tracking; synthetic satellite data; and upscaling games with DeepFakes
What might be the most expensive form of data to gather in the world? View this email in your browser Welcome to Import AI, a newsletter about artificial intelligence. Forward this email to give your
5/18/2020
11 : 7
Import AI 198: TSMC+USA = Chiplomacy; open source Deepfakes; and environmental justice via ML tools
How might generative design influence financial markets - could we imagine a world where a factor in a company's share price is the speed with which it can semi-autonomously generate new products
5/11/2020
11 : 6
Import AI 197: Facebook wants cyberpunk AI; Chinese companies build 'AIBench'; plus, how Cloudflare uses AI
In the same way entertainment companies release games then subsequently DLC, might AI organizations release large models then subsequent DLC-like finetuned versions and/or 'feature plugins'?
5/4/2020
11 : 6
Import AI 196: Baidu wins city surveillance challenge; COVID surveillance drones; and a dataset for building TLDR engines
Job alert: Policy Manager, OpenAI. Happy to chat informally about the role via email. View this email in your browser Welcome to Import AI, a newsletter about artificial intelligence. Forward this
4/27/2020
11 : 37
Import AI 195: StereoSet tests bias in language models; NLP Spy vs Spy; plus, using NLP to waste phishers' time
What might a robot-backed mechanical turk look like, and what sorts of data could you use it to generate? View this email in your browser Welcome to Import AI, a newsletter about artificial
4/20/2020
10 : 7
ImportAI 194: DIY AI drones; Audi releases a self-driving dataset; + Eurovision-style AI pop.
How might a slow internet work - difference engine search engines operating over geared memories? View this email in your browser Welcome to Import AI, a newsletter about artificial intelligence.
4/14/2020
8 : 46
Import AI 193: Facebook simulates itself; compete to make more efficient NLP; face in-painting gets better
if people succeed at creating artificial general intelligence, could we imagine 'demoscene AGI implementations', where people compete to see who can implement an AGI-class system in the