September 26, 2025· 45 min

Jack Morris on Finding the Next Big AI Breakthrough

Orality
Model
50%

Speaker Breakdown

HostJoe Weisenthal(2,628 words)
M:29%
HostTracy Alloway(4,982 words)
M:29%
GuestJack Morris(1,263 words)
M:94%

Oral Indicators

Agonistic32%
literally, completely, obviously
Engagement65%
you, our, your
Memory Aids100%
listen, so, now
Repetition100%
like (275x), they (87x), think (86x)
Parallelism69%
So have you heard the story ab..., So switch to Verizon Business,..., And I'm Tracy Alloway....
Sound Patterns91%
99 question(s), alliteration: "markets move", alliteration: "barclays brief"
Formulaic Phrases6%
at the end of the day, you know what, i mean

Literate Indicators

Hedging5%
could, might, quite
Passive Voice4%
been affiliated, been trained, is called
Abstract Nouns15%
investment, prescription, medication
Subordination6%
because, though, while
Sentence Length29%
Avg: 12.4 words/sentence
Word Complexity45%
investment, analyze, anticipate
Academic Markers0%
Impersonal Style35%
708 personal pronouns found
Descriptive Style100%
automatically, family, apply

Description

We know that the top-tier AI labs are spending unbelievable amounts of money on talent. But what are these researchers actually working on? And how do we know that they're making progress? And furthermore, how can we even measure that progress? On this episode, we speak with Jack Morris, an AI researcher and Ph.D. candidate at Cornell University, who is also a part-time researcher at Meta. We talk about what he does, and why breakthroughs seem to be lumpy and unpredictable. We also talk about the battle between open- and closed-source approaches, US vs. Chinese labs, and how an individual talent thinks about where they want to spend their time, balancing the desire for research and prestige with a big fat paycheck. Only Bloomberg.com subscribers can get the Odd Lots newsletter in their inbox — now delivered every weekday — plus unlimited access to the site and app. Subscribe at bloomberg.com/subscriptions/oddlots See omnystudio.com/listener for privacy information.