Assertion certainty 4/5 debate potential 2/5

Medlock: 2015 AI technology cannot abstract concepts across domains

Ben Medlock · Ben Medlock, SwiftKey // Building a Better Keyboard // Data Driven NYC (FirstMark Capital) · Apr 2, 2015 · at 5:37

Ben Medlock, co-founder and CTO of SwiftKey, discusses the limitations of 2015 artificial intelligence compared to human intelligence at Data Driven NYC.

0:00 / 0:06exact quote · 6.6s

▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark

“This is what the human brain does so incredibly well, and what our technology today is currently almost entirely unable to do.”

quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →

The claim, plainly

Ben Medlock states that AI technology in 2015 is almost entirely unable to acquire concepts from one domain and abstract and apply them to another domain, unlike the human brain.

a quote alone can be ambiguous, so this restates it with the references and context of the conversation around it resolved. AI-written; the tape below the player is the source.

How this resolves

Verification couldn't pin this to a public evidence class, so it stays on the record with its date, unassessed rather than guessed.

prediction date: 2015-04-02, stated by the speaker

Mentioned

AI technology Ben Medlock human brain

⚑ Spot an error? Report it

More from Ben Medlock

Insight

Ben Medlock: Touchscreen typing is an inference problem, not a mechanical one

“This is a virtual keyboard, and the problem of typing was now not a mechanical problem. It's an inference problem, and that means that there is a degree of uncertainty.”

Ben Medlock Apr 2, 2015 ▶ 6:40 Ben Medlock, SwiftKey // Building a Better Keyboard // Data Driven NYC (FirstMark Capital)

Assertion Partly supported

Medlock: SwiftKey is used on close to a billion devices worldwide

“We designed this in 2010, and it's now used on close to a billion devices worldwide.”

Ben Medlock Apr 2, 2015 ▶ 7:42 Ben Medlock, SwiftKey // Building a Better Keyboard // Data Driven NYC (FirstMark Capital)

Disclosure

Medlock: SwiftKey chose independent probability distributions over neural networks for tractability

“And we took the first one of these because it's a lot easier and a lot more tractable.”

Ben Medlock Apr 2, 2015 ▶ 9:44 Ben Medlock, SwiftKey // Building a Better Keyboard // Data Driven NYC (FirstMark Capital)

Insight

Medlock: N-gram models are extremely hard to beat in voice recognition

“If you've done anything in voice recognition, you will know that it's really, really hard to beat Ngram models.”

Ben Medlock Apr 2, 2015 ▶ 11:10 Ben Medlock, SwiftKey // Building a Better Keyboard // Data Driven NYC (FirstMark Capital)

Insight

Medlock: Neural network language models are a highly promising research avenue

“Another possibility is to move away from engrams and towards neural network-based language models, which is a really promising, ah, avenue of research at the moment.”

Ben Medlock Apr 2, 2015 ▶ 11:49 Ben Medlock, SwiftKey // Building a Better Keyboard // Data Driven NYC (FirstMark Capital)

Disclosure

SwiftKey scraped the entire public internet to train its language models

“We scraped all of the publicly available texts off the internet and dumped it into language specific buckets.”

Ben Medlock Apr 2, 2015 ▶ 12:24 Ben Medlock, SwiftKey // Building a Better Keyboard // Data Driven NYC (FirstMark Capital)

Medlock: 2015 AI technology cannot abstract concepts across domains

More from Ben Medlock

Turn any episode into a week of clips.