Messages
10/4/2021
10 : 34
Import AI 268: Replacing ImageNet; Microsoft makes self-modifying malware; and what ImageNet means
If three different species developed computer systems, how different would those systems be? View this email in your browser Welcome to Import AI, a newsletter about artificial intelligence. Forward
9/27/2021
10 : 24
Import AI 267: Tigers VS humans; synthetic voices; agri-robots
What happens when AI assistants become AI directors? How large can a computer get before distances between processors and memory offset scaling gains View this email in your browser Welcome to Import
9/20/2021
11 : 4
Import AI 266: DeepMind looks at toxic language models; how translation systems can pollute the internet; why AI can make local councils better
Given a long enough time period, is it inevitable that a conscious species invents artificial intelligence? Or is high-powered augmentation a plausible evolutionary path? View this email in your
9/13/2021
11 : 34
Import AI 265: Deceiving AI hackers; how Alibaba makes money with ML; why governments should measure AI progress
In a thousand years, what might be the best records of contemporary AI systems circa 2021? Models stored on memory diamond? AI algorithms encoded in DNA? Stone sculptures based on 3D CLIP outputs? View
8/30/2021
9 : 4
Import AI 264: Tracking UAVs; Stanford tries to train big models; deepfakes as the dog ate my homework
A lot of the biggest users of AI are predominantly web entities that deal in the virtual rather than the physical (eg, Google, Facebook). Though there are some applications of modern AI to robots,
8/23/2021
12 : 4
Import AI 263: Foundation models; Amazon improves Alexa; My Little Pony GPT.
If you wanted to compute something over the course of one million years, what would you do? View this email in your browser Welcome to Import AI, a newsletter about artificial intelligence. Forward
8/16/2021
14 : 14
Import AI 262: Israeli GPT3; Korean GLUE; the industrialization of computer vision
What happens when there are too few computers for too few buyers? COVID gave us a taste of this with covid-induced + crypto supply shortages meaning NVIDIA GPUS weren't available. Could this become
8/9/2021
10 : 24
Import AI 261: DeepMind makes a better Transformer; drones can see in the dark now; and a 6bn finetuned code model.
Will we ever be able to harness the full distributed processing capability of the world's smartphones? And if we were able to, what would we do with this? View this email in your browser Welcome to
8/2/2021
10 : 34
Import AI 260: BERT-generated headlines; pre-training comes to RL; $80 million for industrial robots
When might data centers start being built for the purpose of running inference for single, massive AI models? View this email in your browser Welcome to Import AI, a newsletter about artificial
7/26/2021
12 : 4
Import AI 259: Race + Medical Imagery; Baidu takes SuperGLUE crown; AlphaFold and the secrets of life
What is the difference between interpolation and creativity? View this email in your browser Welcome to Import AI, a newsletter about artificial intelligence. Forward this email to give your chums an
7/19/2021
11 : 4
Import AI 258:Game engines are data generators; Spanish language models; the logical end of civilization
If there were three planetary-scale AIs on three different planets with a 15 minute delay between them, then how might a conflict unfold? View this email in your browser Welcome to Import AI, a
7/12/2021
12 : 4
Import AI 257: Firefighting robots; how Europe's AI legislation falls short; what the DoD thinks about responsible AI
Would a dataset of the entire universe be sufficient to encapsulate anything a being stationed in that universe could imagine? Or would it be insufficient in some way? View this email in your browser
7/6/2021
14 : 14
Import AI 256: Facial recognition VS COVID masks; what AI means for warfare; CLIP and AI art
Will computer viruses ever become so complex that we might consider them sentient? View this email in your browser Welcome to Import AI, a newsletter about artificial intelligence. Forward this email
6/21/2021
13 : 4
Import AI 254: Facebook uses AI for copyright enforcement; Google uses RL to design better chips.
Do organizations with significant AI investments make faster decisions (in certain areas) than ones which haven't made these investments? Or is it more that the organizations which invest in AI are
6/14/2021
18 : 4
Import AI 253: The scaling will continue until performance saturates
If certain types of AI progress are predictable, then should the government anticipate certain soon-to-arrive capabilities and alter the behavior of its own institutions? View this email in your
6/7/2021
12 : 4
Import AI 252: Gait surveillance; a billion Danish words; DeepMind makes phone-using agents
As off-the-shelf AI advances, the potential for emergence increases; in a few years, perhaps some of our most impactful AI systems will be assembled at home by hobbyists out of pre-built components.
5/31/2021
11 : 4
Import AI 251: Korean GPT-3; facial recognition industrialization; faking fingerprints with GANs
Plus: a short story about a robot imagining another robot Will the 'personality' types of AI systems be more or less varied than the types of the people that create them? View this email in
5/24/2021
11 : 34
Import AI 250: Facebook's TPU; Twitter analyzes its systems for bias; encouraging proof about federated learning
Is symbolic AI the 'dark matter' of AI - there's tons of it deployed around us and we can't measure it. Or is it far more insubstantial? And how could we know the truth? View this email
5/17/2021
10 : 34
Import AI 249: IBM's massive code dataset; dataset archaeology: BookCorpus; Facebook wants computers to read the world
There are more than six thousand languages used in the world today - how many languages might AI systems evolve to communicate with another? View this email in your browser Welcome to Import AI, a
5/3/2021
11 : 4
Import AI 247: China makes its own GPT3; the AI hackers have arrived; four fallacies in AI research.
How might different alien intelligences conceive of AI? If - or perhaps when - we meet aliens, will they have also developed things that seem like neural networks? How much diversity is possible in the