A.I. is at it again

[pbmax]

Is he? Humans are wrong on their facts all the time. I probably have my facts wrong ten times a day. Then I Google (which uses AI) or use a dedicated AI tool to find the correct facts.

If you rely on the AI influenced results your error rate won’t improve.

It can’t discern who the opponents were in the 1982 Word Series even though that data exists in multiple sources.

The error is in assuming either is flawless.

[sharpe1027]

If you rely on the AI influenced results your error rate won’t improve.

It can’t discern who the opponents were in the 1982 Word Series even though that data exists in multiple sources.

The error is in assuming either is flawless.

Yes, it will. Neither is flawless, but AI with a human check or human with an AI check is better than either alone.

It's only getting better too. If your basing your opinion on free tools or what you saw a few months ago, you're already way behind the curve.

[sharpe1027]

It can’t discern who the opponents were in the 1982 Word Series even though that data exists in multiple sources.

Actually, the 1982 World Series (Cardinals vs. Brewers) is trivial for most modern LLMs to retrieve correctly, so that specific example is likely from an outdated LLM or a fluke of a poorly structured query.
AI models do still hallucinate and hit reasoning walls independently of how well they're prompted.

The real shift is recognizing that AI isn't a human encyclopedia. it's a probabilistic prediction engine. It fails in weird, non-human ways, but it also handles synthesis and generation at a scale humans can't touch. It's less about AI errors and more about learning how to use a fundamentally different kind of tool.

[pbmax]

Actually, the 1982 World Series (Cardinals vs. Brewers) is trivial for most modern LLMs to retrieve correctly, so that specific example is likely from an outdated LLM or a fluke of a poorly structured query.
AI models do still hallucinate and hit reasoning walls independently of how well they're prompted.

The real shift is recognizing that AI isn't a human encyclopedia. it's a probabilistic prediction engine. It fails in weird, non-human ways, but it also handles synthesis and generation at a scale humans can't touch. It's less about AI errors and more about learning how to use a fundamentally different kind of tool.

It sounds like it was the free Gemini. But I would like my old search back while we are at it.

[bobblehead]

Is he? Humans are wrong on their facts all the time. I probably have my facts wrong ten times a day. Then I Google (which uses AI) or use a dedicated AI tool to find the correct facts.

Yes...me too. Usually before I post because I hate looking stupid. Not everyone seems bothered by that though.

[bobblehead]

It sounds like it was the free Gemini. But I would like my old search back while we are at it.

I just used gemini to ask who the leading hitter in the 1982 world series was....not who were the 2 participants. Here is the detailed answer I got in about 2 seconds. No human could have done this:


Leading Hitter of the 1982 World Series


Darrell Porter of the St. Louis Cardinals was the leading hitter in the 1982 World Series, earning the Most Valuable Player award Wikipedia+1.

Porter, a catcher for St. Louis, batted .286 with five RBIs, including a home run in Game 4 grokipedia.com. His contributions were pivotal in the Cardinals’ 4–3 series win over the Milwaukee Brewers, which gave St. Louis its first championship since 1967 Wikipedia.

While other notable hitters emerged during the series — such as Paul Molitor, who set a World Series record with 11 hits over the seven games grokipedia.com, and Keith Hernandez, who delivered key hits and RBIs in Game 6 — Porter’s combination of on-base skills, timely hitting, and defensive play earned him the MVP designation Wikipedia+1.

Key 1982 World Series hitting highlights:

• Darrell Porter (STL) – .286 AVG, 5 RBI, HR in Game 4, MVP Wikipedia+1
• Paul Molitor (MIL) – 11 hits (MLB record for a 7‑game series), 5 hits in Game 1 grokipedia.com
• Keith Hernandez (STL) – 4 RBI in Game 6, decisive hit in Game 7 grokipedia.com

In short, Porter was both the leading hitter and the MVP of the 1982 World Series, making him the standout offensive performer of the championship.

That said, its not perfect by any means. Robin Yount actually had 12 hits and hit .414 which made him the actual best hitter by that metric and Molitor wasn't the hit leader in that series. I got that from Grok, so the above info is extensive and wrong on one account at least, so its not perfect, but again, using 2 agents got me a lot of information I couldn't have gotten on my own without spending about 20 minutes looking.

[pbmax]

That's why I would like regular search back. That summary is gobbledygook in an attempt to answer your question. Its fast as opposed to discerning or careful. Its pattern recognition. It looks for things associated with "leading hitter" and comes back with the 2nd most hits on the Cardinals, tied with 3 others for second most RBIs and doesn't consider Yount (with the most hits in the series and 1 more RBI than Porter).

Its almost useless unless you are prepared to go in and do the dive yourself. Or do multiple AI searches. Or just read Baseball Reference and Wikipedia.

The benefit of actually reading is all the tangential stuff and context you will pick up. Gemini is like an unreliable Cliff Notes.

[sharpe1027]

It sounds like it was the free Gemini. But I would like my old search back while we are at it.

AI tools can show you several options to get your search results without AI.

[sharpe1027]

Depending on the statistical category used to define the "leading hitter," the leaders in the 1982 World Series were: * Overall Batting Average, Slugging, & OPS (min. 15 AB): Dane Iorg (St. Louis Cardinals) led all players with a .529 batting average (9-for-17), .529 OBP, .882 SLG, and 1.412 OPS across 5 games as a designated hitter.
* Total Hits & Full-Time Starter Batting Average: Robin Yount (Milwaukee Brewers) led all players with 12 hits (12-for-29) and led all everyday starters who played in all 7 games with a .414 batting average (.452 OBP, .621 SLG, 1.072 OPS). (Paul Molitor was second in total hits with 11, batting .355 over 31 at-bats).
* Home Runs: Willie McGee (Cardinals) and Ted Simmons (Brewers) tied for the series lead with 2 home runs each.
* Runs Batted In (RBI): Keith Hernandez (Cardinals) led the series with 8 RBIs.
* World Series MVP: Darrell Porter (Cardinals), who batted .286 (8-for-28) with 1 home run and 5 RBIs.

[sharpe1027]

Gemini Argon, asking for a concise answer with nothing made up

[pbmax]

AI tools can show you several options to get your search results without AI.

Yeah, I usually scroll down to the typical results but my understanding is that even though results are influenced by rankings influenced by the AI rather than the old school rankings.

[sharpe1027]

Yeah, I usually scroll down to the typical results but my understanding is that even though results are influenced by rankings influenced by the AI rather than the old school rankings.

Fair point, though "old school" rankings have been dead for a long time. Google started baking machine learning into the core link rankings back in 2015 with RankBrain. It's been a constantly mutating algorithm since the beginning—the real difference now is just generative summaries at the top vs. neural-network ranking at the bottom. Both are AI, just doing different jobs.

This response was supplemented by AI.

[bobblehead]

That's why I would like regular search back. That summary is gobbledygook in an attempt to answer your question. Its fast as opposed to discerning or careful. Its pattern recognition. It looks for things associated with "leading hitter" and comes back with the 2nd most hits on the Cardinals, tied with 3 others for second most RBIs and doesn't consider Yount (with the most hits in the series and 1 more RBI than Porter).

Its almost useless unless you are prepared to go in and do the dive yourself. Or do multiple AI searches. Or just read Baseball Reference and Wikipedia.

The benefit of actually reading is all the tangential stuff and context you will pick up. Gemini is like an unreliable Cliff Notes.

But be fair. I asked a subjective question. When searching for nothing but facts AI is really helpful. I don't miss the old searches because they were brutally politically biased. Not violate rules here, but there were a lot of question the old searches just simply would return so biased that it was sick or often wouldn't even give me legitimate links. Now if I want to know a lot of simple facts I can get them quickly. And Gemini usually gives you the "old search" links if you scroll past the summary.

[bobblehead]

Depending on the statistical category used to define the "leading hitter," the leaders in the 1982 World Series were: * Overall Batting Average, Slugging, & OPS (min. 15 AB): Dane Iorg (St. Louis Cardinals) led all players with a .529 batting average (9-for-17), .529 OBP, .882 SLG, and 1.412 OPS across 5 games as a designated hitter.
* Total Hits & Full-Time Starter Batting Average: Robin Yount (Milwaukee Brewers) led all players with 12 hits (12-for-29) and led all everyday starters who played in all 7 games with a .414 batting average (.452 OBP, .621 SLG, 1.072 OPS). (Paul Molitor was second in total hits with 11, batting .355 over 31 at-bats).
* Home Runs: Willie McGee (Cardinals) and Ted Simmons (Brewers) tied for the series lead with 2 home runs each.
* Runs Batted In (RBI): Keith Hernandez (Cardinals) led the series with 8 RBIs.
* World Series MVP: Darrell Porter (Cardinals), who batted .286 (8-for-28) with 1 home run and 5 RBIs.

And that is sort of my point. I asked a subjective question. If I asked it in a thread I'd get 10 different answers from ten posters. But asking for a fact (most hits) it generally nails it down.

[sharpe1027]

And that is sort of my point. I asked a subjective question. If I asked it in a thread I'd get 10 different answers from ten posters. But asking for a fact (most hits) it generally nails it down.

It actually does fine with subjective questions.That’s often where it shines. By asking for nuance rather than a single metric, it broke down the top potential candidates across multiple categories in just a few seconds.