Assertion Supported AI assessment confidence: 95% certainty 4/5 debate potential 1/5

Zaremba: Google avoided ML in Search early on over interpretability issues

Wojciech Zaremba · An AI Primer with Wojciech Zaremba · Y Combinator · May 17, 2017 · at 11:53

OpenAI co-founder Wojciech Zaremba explains how AI differs from ML, noting Google Search's evolution from algorithmic PageRank to machine learning models.

0:00 / 0:21exact quote · 22.0s

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

“Over the time Google search started to use machine learning because it was, it helps to improve results but simultaneously, they wanted to avoid it for some time as it's more difficult to interpret the results, and it's more difficult to actually understand what system does.”

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

Google initially avoided using machine learning in Search because interpreting results and understanding the system's behavior was difficult.

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.

Assessment

Historical accounts and statements from Google engineering leadership confirm that Google intentionally avoided relying on machine learning for core search ranking for many years due to the difficulty of interpreting, explaining, and debugging black-box models. Peter Norvig (then Director of Research) in 2008, Edmond Lau (former Search Quality engineer) in 2011, and Jeff Dean in 2016 all documented that the Search team, under Amit Singhal, favored rule-based heuristics and manually crafted scoring formulas because they allowed engineers to directly diagnose and control ranking behavior. It was not until the rollout of RankBrain in late 2015 that machine learning was deeply integrated into core search ranking.

In May 2008, then-Director of Research Peter Norvig explained that Google avoided purely machine-learned models for search ranking to prevent catastrophic errors on novel queries and maintain direct human control.
In April 2011, former Search Quality engineer Edmond Lau detailed that Google's core ranking team avoided machine learning because ML models are 'black boxes' that make it difficult to explain why one result ranked above another or directly debug edge cases.
In June 2016, Steven Levy and Jeff Dean documented that the Search Ranking Team resisted deep neural networks until sufficient interpretability and debugging mechanisms could be established.
Google did not deploy machine learning broadly in core ranking until RankBrain was launched in late 2015.

Checked against

[1] forbes.com [2] wikipedia.org [3] ethanhays.org [4] quora.com [5] quora.com [6] quora.com [7] quora.com [8] umn.edu

"RankBrain" "Steven Levy" "Wired" "In a machine learning system, it's hard to explain and ascertain why a particular search result ranks" "Anand Rajaraman" "Peter Norvig" "machine learned" search "machine learning" "Singhal" "black box" search ranking "machine learning" "Amit Singhal" "Wired" OR "Bloomberg" "RankBrain" Amit Singhal machine learning search debug In The Plex "machine learning" "Amit Singhal" "In the Plex" "Edmond Lau" "machine learning" "search ranking" "How Google Is Remaking Itself as a Machine Learning First Company" "Singhal" OR "ranking" "Steven Levy" "How Google Is Remaking Itself as a Machine Learning First Company" search

assessed as of 2026-09-05

This assessment is an AI reading of a public statement, made against the cited references with the confidence shown. It can be wrong. The quote, the tape and the references are all on this page; if something is off, use report an error below. full disclaimer →

Mentioned

Google Google Search machine learning

⚑ Spot an error? Report it

More from Wojciech Zaremba

Opinion

Zaremba: Universal basic income is the only viable path as automation grows

“So, I believe that we'll have to offer to people a basic income. I super strongly believe that actually that's the only way.”

Wojciech Zaremba May 17, 2017 ▶ 52:06 An AI Primer with Wojciech Zaremba · Y Combinator

Insight

Zaremba: Most business problems can be solved via supervised learning

“Majority of business problems can be framed as supervised learning, and therefore they can be solved with current techniques, as long as we have sufficient number of input examples and what we want to predict”

Wojciech Zaremba May 17, 2017 ▶ 46:59 An AI Primer with Wojciech Zaremba · Y Combinator

Assertion Partly supported

Zaremba: OpenAI has secured $1 billion in total investment

“In total we gather an investment of one billion dollar in the group.”

Wojciech Zaremba May 17, 2017 ▶ 1:34 An AI Primer with Wojciech Zaremba · Y Combinator

Insight

Zaremba: Reinforcement learning struggles in reality due to reward and reset assumptions

“The assumption underlying reinforcement learning Is that then there is some environment, and environment, you are an agent, and you are acting in environment by executing actions and getting rewards from the environment. And the rewards might be taught as, let…”

Wojciech Zaremba May 17, 2017 ▶ 7:12 An AI Primer with Wojciech Zaremba · Y Combinator

Opinion

Zaremba: ImageNet is the essential dataset that enabled deep learning

“That's the essential data set that made deep learning happen.”

Wojciech Zaremba May 17, 2017 ▶ 34:36 An AI Primer with Wojciech Zaremba · Y Combinator

Assertion Supported

Zaremba: ImageNet error dropped to 3%, achieving superhuman vision performance

“Within several years, people got down, I believe, to three percent error, and that's essentially superhuman performance.”

Wojciech Zaremba May 17, 2017 ▶ 37:40 An AI Primer with Wojciech Zaremba · Y Combinator

Zaremba: Google avoided ML in Search early on over interpretability issues

More from Wojciech Zaremba

Turn any episode into a week of clips.