There are Bad Algorithms in our future, points out Algorithms that are biased. Algorithms that quickly learn bad behavior. Algorithms that are evil by design. Algorithms that we can no longer really control.
timharford.com/2019/03/just-b
Conversation
Replying to
The age of algorithm will have some serious problems. Imagine the truck terrorists getting their hands on adversarial AI techniques
Quote Tweet
Woah.
Researchers successfully trick @Tesla autopilot into driving into opposing traffic via “small stickers as interference patches on the ground” #adversarialAI #abusabilitytesting (ht @ryanaraine @keen_lab)
keenlab.tencent.com/en/whitepapers
Show this thread
1
1
Replying to
Forget Black Swans. Machine learning algorithms are bad at dealing with even light grey swans
Quote Tweet
Replying to @SuB8u
This is fascinating. #Brexit is a pain for algorithms!
"Machine-driven trading systems in the $5.1 trillion-a-day global currency market are struggling to cope with the blizzard of headlines about Britain’s efforts to extricate itself from the EU"
uk.reuters.com/article/us-ste
2
1
Replying to
The age of algorithms taking over (badly) is beginning. Thread:
Quote Tweet
It's a big day today: I had a scientific manuscript rejected by a robot. Thread. The bot detected ❝a high level of textual overlap with previous literature❞. In other words, plagiarism. 1/6
Show this thread
1
1
4
Replying to
For all those who think machine learning is easy/cheap, please note the costs of training a deep learning network
Quote Tweet
Holy crap: It costs $245,000 to train the XLNet model (the one that's beating BERT on NLP tasks..512 TPU v3 chips * 2.5 days * $8 a TPU) - arxiv.org/abs/1906.08237
Show this thread
2
1
3
Replying to
As the US Supreme Court indicates that access to mobile phone records requires warrants, asks this question:
do we want our phones to have greater protections than our faces?
In this week's newsletter dedicated to #surveillance #biometrics
1
3
2
Replying to
“Software engineers with no people skills are teaching our future robots people skills”
Quote Tweet
Is anyone else worried that software engineers with no people skills are teaching our future robots people skills
Show this thread
3
1
5
Replying to
Beware: 90% of AI/ML today is in this category, where "seed is also a hyper-parameter"
2
1
3
Replying to
What?! This dress tricks the AI into believing that you are a car!
Quote Tweet
Replying to @SuB8u
Love this. Adversarial fashion is here. Clothing that tricks surveillance algorithms into believing you are a car!
cnet.com/news/anti-surv
1
4
9
Replying to
Machine learning see, machine learning do
Quote Tweet
Beware causal inference with machine learning because the answer isn’t in the training data, and the algorithm can only learn what it sees.
#causalinference #ML #AI
twitter.com/uzunkokcan/sta
2
19
31
Replying to
1
1
Question to ask is - does imitative behaviour ever end and people actually start thinking for themselves?
Answer is no.
1
2
Show replies









