Undated
Some surveyed companies reported that no use of generative AI was authorized, which Davenport considers a bad idea.
“there was no authorized use of generative AI in their companies. I think in general, that's a bad idea”
Topic
Generative AI
13 facts from 4 episodes, in date order. These are what speakers said on the shows, each linked to the moment it was said. They aren't independently verified.
Undated
In the survey Davenport cited, only eight percent of organizations said they had generative AI systems in production.
“only eight of organizations said they had systems like this in production.”
Undated
Prem Natarajan says the technical term for generative AI models following prompts and instructions without explicit training for the task is in-context learning.
“And if you become really technical, they'll call it in context learning.”
Undated
Prem Natarajan names retrieval augmented generation as one of the most popular ways generative AI is currently used, including in search tools.
“one of the most popular ways in which these are being used are this retrieval augmented generation style use.”
Undated
Rao argued that generative image models misspell text because they treat letters as symbols, which he cited as evidence they don't copy the images they were trained on.
“it always gets text wrong, generative AI always gets text wrong, cause it has no idea it's words.”
Undated
Barney says firms are using generative AI to produce first drafts of code, which humans then refine.
“Firms are using GenAI to, right now, produce the first draft of code, and then a human being will go through and then update and make it make it more elegant or whatever.”
Undated
Barney says boilerplate contracting work in the legal profession is expected to be substantially affected by generative AI.
“in the legal profession, a lot of the sort of boilerplate contracting work and and other kinds of work that the lawyers historically have done is”
Undated
Barney says sampling theory suggests seeking the smallest dataset that reveals the patterns needed, and that research is under way on minimal datasets for generative AI.
“sampling theory suggests that, uh, we should actually be trying to figure out what's the smallest sample of smallest data set we need to have in order to generalize”
Undated
Barney says results from using generative AI feed back into its data, which creates a second-mover advantage.
“when you use GenAI to do something, the results of the analysis you've done then feed back into the data set, and so the GenAI becomes smarter after you've done your work.”
Undated
Barney says companies should use generative AI widely and quickly because access is currently inexpensive.
“Obviously, you should be using GenAI pretty much everywhere where you can, as quickly as you can, especially since access to GenAI is actually not that expensive right now.”
Undated
The Barney and Reeves paper uses the example of asking generative AI for seven new kinds of toothbrushes to illustrate work that yields only competitive parity.
“We use the example in the, uh, in the paper of, uh, toothbrush example, you know, asking GenAI to generate a list of 7 new kinds of toothbrushes.”
Undated
Barney illustrates that a firm of 10,000 employees might run about 20,000 generative AI experiments a day.
“we have 10,000 employees, that means there are probably 20,000 GenAI experiments a day, some of which have no value at all, but some of which might be quite valuable”
Undated
Barney suggests companies run generative AI competitions, with prizes such as a thousand dollars, to harness employee experimentation.
“you can have GenAI competitions where you say, Hey, if you come up a really cool application of GenAI, submit it to this people, this this group”
Episodes