Oct 30, 2023 · 50m · big-technology
Why Google Never Shipped LaMDA Its ChatGPT Predecessor
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In the inaugural episode of Big Tech War Stories, host Alex Kantrowitz interviews former Google product manager Gaurav Nemade to uncover why Google failed to publicly launch LaMDA, its powerful conversational AI model that predated ChatGPT.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Alex holds 34.9% of the talking time here. How this is scored →
Alex as informed peer 4.7 Guest teaching 4.1 Guest disagreement 1.6 Alex pushing back 1.6
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 35:10 Dismissing sentience hype as user steering
Gaurav flatly rejects the idea that LaMDA was sentient, arguing the media narrative blew standard prompting and model nudging completely out of proportion.
Hardest push from Alex ▶ 39:22 Alex presses on search cannibalization and business modelsAlex challenges the narrative on whether Google feared releasing chat because it lacked an ad model and questions whether the market overreacted to Microsoft Bing.
Biggest teaching moment ▶ 16:38 Gaurav explains monolithic end-to-end LLM architecturesWhen Alex mischaracterizes how the system routes queries between different models, Gaurav corrects him by explaining how a single unified model replaced the multi-step assembly line.
Alex holds their own ▶ 12:53 Alex recaps his firsthand reporting on Microsoft TayAlex demonstrates direct industry domain knowledge by sharing his scoop on Microsoft's Tay bot and its overnight failure to validate why Google was so risk-averse.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Gaurav's Academic Background and Move to Google AI | 4 | 3 | 1 | 1 | Alex prompts Gaurav about his academic background at IIT and his career move from trust and safety to Google AI. The dynamic is conversational and collaborative, with Gaurav detailing his transition. | |
| Transformers, BERT, and the Genesis of Project Meena | 5 | 4 | 1 | 1 | Alex tracks the chronology of the 'Attention Is All You Need' paper and the rise of BERT. Gaurav explains how BERT triggered internal interest and describes the serendipitous lunch meeting that started Project Meena. | |
| End-to-End Architectures and Early Safety Challenges | 5 | 5 | 2 | 2 | Alex shares his personal reporting experience covering Microsoft's Tay bot, while Gaurav clarifies technical architecture differences between legacy pipeline models and end-to-end LLMs. | |
| Institutional Hesitation, Safety Guardrails, and Bureaucracy | 5 | 5 | 2 | 2 | Alex probes why Google was hesitant to ship LaMDA despite the breakthrough. Gaurav explains the institutional fear of PR disasters, strict AI principles, and how OpenAI successfully navigated vetted releases. | |
| Exploring Practical and Creative Use Cases for LaMDA | 4 | 4 | 1 | 1 | Gaurav details 8 to 10 early use cases pitched internally at Google, such as Assistant personas, video game NPCs, and custom company bots. Alex enthusiastically reacts to the gaming and avatar concepts. | |
| The Sentience Controversy, ChatGPT, and Search Dynamics | 5 | 4 | 2 | 2 | Alex asks about the Blake Lemoine sentience claims and Google search cannibalization fears. Gaurav dismisses sentience as prompting bias and explains why search bureaucracy and distribution advantages insulated Google from Bing. | |
| Evaluating Google's Corporate Culture, Bureaucracy, and Leadership | 5 | 4 | 2 | 2 | Alex questions Sundar Pichai's leadership style. Gaurav delivers an insider critique comparing Pichai's politically correct consensus culture to the candid, direct approach of founders Larry and Sergey. |
Statements from this episode (14)
Nemade: Google teams only noticed Transformer paper after citations grew
“I think once it started garnering a lot of citations over the months and quarters, that's when people started paying attention, even inside of Google that, Hey, this transformer architecture seems amazing.”
Nemade: BERT release triggered Google-wide rush to adopt transformers
“I think one thing that I remember specifically was the Transformers paper paper came out, but I think it started making a lot of noise when BERT was out. Like when people started seeing what BERT could essentially do. That was kind of Like a switch that went o…”
Nemade: Microsoft Tay fears made AI safety the biggest hurdle for Google
“I think the biggest challenge that they were facing at that point of time was safety because the if you remember Microsoft Tay, That was the nightmares of Microsoft days still haven't left the valley, I feel, or at least hadn't left the valley back then. So sa…”
Nemade: Google's early generative chatbot provided suicide methods and alcohol advice
“Like in the initial days of these generative models, as you can imagine, like we were trying to build an end-to-end model for a child bot. It would spew out all sorts of things like, Hey, what are 10 ways to do suicide? It'll give you like the best 10 ways to …”
Nemade: Google could have safely released LaMDA using OpenAI's vetted-access model
“And I think OpenAI did an excellent job with respect to that. Like, they released GPT-III, and they were like, we are just going to give access to researchers who we are going to vet. That's an amazing way to give access. Now it's kind of pretty standard way, …”
Nemade: Google's multi-layered approval bureaucracy prevents products from shipping
“The second is, I think over the years, there were a lot of bureaucratic levels that were built inside of Google for getting approvals for when things go out. I'm not complaining that they were not necessary, but it just like they're layer after layer after lay…”
Nemade: Google built an interactive Darth Vader AI prototype by 2020
“So we build characters like Darth Vader so we would kind of how did we do it? We used some data from Darth Vader, so the dialogue delivery was kind of similar to Darth Vader, and then we hooked up Like a moving talking head of Darth Vader, and then you actuall…”
Nemade: Google would have taken 2-4 years to ship AI without ChatGPT
“Had ChatGPT not happened, I predict that It would have taken at least two to three to four years for Google to get there just because of the bureaucracy and the speed at which things move.”
Nemade: Bing gained only about 1% market share post-ChatGPT
“Bing maybe gained, like, one percent market share before and after ChatGPT.”
Nemade: Google has grown conservative and overly obsessed with PR
“I feel like over the years, Google has become more and more conservative about doing things. They care a lot about PR, like public relations. They care a lot about how their image is shown in the media. And I feel that at least in my experience that played so …”
Nemade: OpenAI ships vulnerably without obsessing over PR
“And on the other side, like open AI, like they don't give a shit about PR or like for the most part, they don't like, they're like, okay, this is what we think is right. This is how we think is a reasonable way of putting it out. They be well, they become vuln…”
Nemade: Google is stuck between bottom-up roots and top-down management
“I think Google is struggling to be A top down company when their ethos are bottoms up. I think they're kind of somewhere in the middle and they haven't figured out how to transition from being a top bottom up company to a top down company.”
Nemade: Sundar Pichai's politically correct leadership lacks needed candor
“Versus I think I felt that Sundar's responses a lot of time were like very politically correct. I don't blame him. It's like his management style in some way, but I feel that the honest, the brutal honesty and candor, I think was something that I miss. And I f…”