What podcasts say about large language models
Every statement, with the speaker, the exact quote and the moment it was said.
What experts have said about large language models
27 statements · 12 positive · 12 negative · 1 mixed · 2 neutral
Current large language models are highly inefficient in training and serving.
“these models are very inefficient”
Open the episode · What VCs Really Think About Personal AI Agents | E2347Listen at 14:59
Large language models are missing an expertise and context layer.
“it's that vital context that so many LLMs are missing.”
Open the episode · Jason’s Put a $5K Bounty on His Dream Chrome Extension | E2338Listen at 27:50
Large-language-model research can transfer to recommendation systems.
“a lot of the research that's being done in the space, in the large language model space, can port to recommendation systems”
Open the episode · Adam Foroughi, Applovin CEO: Surviving a 92% Drawdown, Ads as ML 1.0 & the $50B Game Ad MarketListen at 4:03
Current concern about large language model harms is insufficient.
“There's not enough concern about the actual harms of large language models.”
Open the episode · The Great AI Debate: Is Artificial Intelligence an Extinction Threat? Debating the True Risks of Advanced ModelsListen at 3:55
Civilization should not rely on large language models ceasing to improve.
“I don't think we should bet civilisation on the LLMs running out of steam.”
Open the episode · AI Debate Ed Zitron, Andrew McAfee, Nate Soares, Roman YampolskiyListen at 39:17
Training on human text can produce AI systems potentially smarter than humans.
“training AIs to predict human text is training them to be potentially smarter than the humans.”
Open the episode · AI Debate Ed Zitron, Andrew McAfee, Nate Soares, Roman YampolskiyListen at 1:12:34
Unrestrained LLMs connected to infrastructure could cause a power-system outage.
“Do I think that unrestrained LLM use connected to massive amounts of infrastructure could lead to actually a power system going down? Absolutely.”
Open the episode · AI Debate Ed Zitron, Andrew McAfee, Nate Soares, Roman YampolskiyListen at 2:20:58
Large language models are unlikely to cause the discussed existential outcome.
“I don't think they will lead to what you're talking about.”
Open the episode · The Great AI Debate: Is Artificial Intelligence an Extinction Threat? Debating the True Risks of Advanced ModelsListen at 2:22:34
Current LLMs can create digital proxies indistinguishable from people using their external records.
“the type of models that are running right now in LLM will be able to integrate that information and come up with a digital proxy, a simulacrum that will be indistinguishable from us or from them.”
Open the episode · Martine Rothblatt: Growing Unlimited Organs, AI Consciousness, & Digital Personhood by 2030Listen at 34:44
Current large language models lack basal ganglia and biological reinforcement learning.
“these large language models don't have basal ganglia. They don't use reinforcement learning.”
Open the episode · Most Replayed Moment: Brain Rot Experts - How To Use AI In A Healthy Way!Listen at 14:13
Large language models encode human behavior and interaction as a behavioral modality.
“Human behavior and human interaction is a modality.”
Open the episode · Are AI Agents forming "civilizations" or is this just a psy op? | 2332Listen at 15:51
LLMs dramatically accelerate searching and analyzing annual reports.
“LLM is like that on steroids.”
Open the episode · Andrew Page on small caps, AI & changing his mind on BitcoinListen at 18:53
Large language models can write adequately but not exceptionally.
“right now it can write good, but it can't write great.”
Open the episode · Mexican Batman, Britain’s Downfall, Mr Bean’s Comeback & Jimmy Carr - Rabbit Hole #5 - #1143Listen at 1:33:31
Models trained on averages tend not to take creative risks.
“When you have something that's trained on the averages, it doesn't take risks.”
Open the episode · Mexican Batman, Britain’s Downfall, Mr Bean’s Comeback & Jimmy Carr - Rabbit Hole #5 - #1143Listen at 1:33:41
LLMs are particularly useful for troubleshooting from technical logs.
“The troubleshooting thing is awesome for it. It's the. The one weakness I have. It's like genuinely being able to drop a log into it. That's awesome.”
Open the episode · The Man Who Calls BS On AI: AI Is The World’s Greatest SCAM, And They All Know It! | Ed ZitronListen at 1:31:30
LLMs tend toward average, probable outputs rather than personally appropriate results.
“LLMs are not good at this. They regress to the mean and they find the most probable thing.”
Open the episode · Neurosymbolic AI outperforms chatbots and product search | E2327Listen at 0:00
Current large language model AI may have limited improvement potential.
“There isn't necessarily... an unlimited capacity of that kind of artificial intelligence to continue to improve”
Open the episode · Top Scientist REVEALS: We Invented Heaven, God Is A Human Invention! | Prof Brian GreeneListen at 1:31:41
Scaling current large language models may not produce unlimited improvement.
“There isn't necessarily an unlimited capacity of that kind of artificial intelligence to continue to improve.”
Open the episode · Top Scientist REVEALS: A Kid In The 29th Century Could Have Made This World! | Prof Brian GreeneListen at 1:31:41
Massive training datasets enabled current AI systems to recognize contextual patterns reliably.
“This current era when the huge data that these algorithms have learned, let's take Gemini or GPT have learned, really created the capability in the machine's learned space.”
Open the episode · Using AI to Increase Your Intelligence & Enrich Humanity | Dr. Fei-Fei LiListen at 26:51
Persistent LLM memory can produce self-reinforcing feedback and bias.
“there's a lot of bias built into that not just by the people who design the systems but by the force of memory function across different chats and threads. You're going to get a lot of self reinforcing feedback.”
Open the episode · #877: Q&A with Tim — The Art of Male Friendship, Mini-Retirements, Higher-Resolution Living, Reinvention in The Age of AI, and MoreListen at 11:18
Large language models could reduce political information asymmetry more than other technologies.
“The thing that I think will save us as a world more than anything else in terms of information availability and reducing the information asymmetry as it applies to politics, are large language models.”
Open the episode · Mark Cuban on the AI Bubble: Who Actually Gets Wiped Out?Listen at 27:36
LLMs are highly useful for analyzing personalized health information.
“From a holistic health perspective. By holistic, I mean having enough data related to medications, supplements, predispositions, side effects, what happened to me two weeks ago, the LLMs have been incredibly helpful.”
Open the episode · #875: The Random Show — Tim and Kevin Talk Retreats, Mortality, AI Predictions, Supplements, Rock Climbing at (Almost) 50, and Not Waiting for “Someday”Listen at 1:08:32
Large language models are expected to become superhuman explainers.
“we also expect these large language models to be superhuman explainers”
Open the episode · Adam Brown – Einstein's happiest thought: General Relativity from scratchListen at 1:36:19
Future language models will make difficult mathematical proofs easier for humans to understand.
“They will be able to take proofs that are difficult to understand and make them easy to understand.”
Open the episode · Adam Brown – Einstein's happiest thought: General Relativity from scratchListen at 1:36:36
Language-model-generated mathematical ideas can be human-interpretable and reusable in new proofs.
“it came up with a very human interpretable idea and then humans were able to fully comprehend it and comprehend it so well they were able to deploy it in a new scenario”
Open the episode · Adam Brown – Einstein's happiest thought: General Relativity from scratchListen at 1:37:16
Language models’ extreme patience makes them useful for exploring unlikely mathematical approaches.
“another aspect of large language models that makes me pretty optimistic, is that they just have extreme patience”
Open the episode · Adam Brown – Einstein's happiest thought: General Relativity from scratchListen at 1:38:04
Tokenization and model processing can lose meaning during language generation.
“A lot of meaning can get lost.”
Open the episode · Inside The Viral Words That Make You Click - Etymology Nerd - #1086Listen at 1:33:29
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.