4 ms·

I would probably make the same call, there is no universe where "Google's Gemini Beloved By All As Impartial Election Assistant" is the headline. It doesn't eve

by Spivak 3y ago

I would probably make the same call, there is no universe where "Google's Gemini Beloved By All As Impartial Election Assistant" is the headline. It doesn't even matter what it says, having it at all is going to be a media shitstorm with accusations of alignment pushing narratives blah blah. The even smarter play would be to shut down the user-facing service for the 3 months leading to the election.

Me1000 3y ago
Exactly. This is a damned if you do damned if you don't situation. LLMs don't have predictable responses, elections can change very quickly, and it's important for Google's brand that Gemini doesn't just start making things up or speak with authority on outdated information. This is a completely reasonable safeguard against a completely foreseeable headline "Gemini agrees that [ELECTION CONSPIRACY THEORY] is true".
- CrimsonCape 3y ago
  Wouldn't a better approach be to build AI that starts from a groundwork of intelligence? It's in the name. Build a system that can answer easy questions correctly, and be judged on % of correct completions before moving to difficult questions. Tiktok, modern food, LLMs... all sugary colorful substitutes for the real thing. We are supposed to be moving away from this empty stuff, except for major corporations that are just as guilty as the teenage perpetual Tiktok browser.
  - Me1000 3y ago
    I would argue that Google didn't want to build this product at all and that it was a market reaction to ChatGPT's success (which suffers from the exact same problem I described above). The folks at DeepMind seemed quite keen on keeping this stuff in their research labs. Whether or not you want to call an LLM an "AI" or not is really immaterial though, and more often than not just a distraction. Users have found conversational answer engine products like ChatGPT and Gemini valuable. They don't have to be perfect to provide value, but their imperfections are ripe for tarnishing the brands of long well established companies. But LLMs also have uses outside of a conversational answer engine; I for one am excited to watch them deployed in games, and also writing unit tests for code.

by Spivak 3y ago

Me1000 3y ago

Exactly. This is a damned if you do damned if you don't situation. LLMs don't have predictable responses, elections can change very quickly, and it's important for Google's brand that Gemini doesn't just start making things up or speak with authority on outdated information. This is a completely reasonable safeguard against a completely foreseeable headline "Gemini agrees that [ELECTION CONSPIRACY THEORY] is true".

CrimsonCape 3y ago
Wouldn't a better approach be to build AI that starts from a groundwork of intelligence? It's in the name. Build a system that can answer easy questions correctly, and be judged on % of correct completions before moving to difficult questions. Tiktok, modern food, LLMs... all sugary colorful substitutes for the real thing. We are supposed to be moving away from this empty stuff, except for major corporations that are just as guilty as the teenage perpetual Tiktok browser.
- Me1000 3y ago
  I would argue that Google didn't want to build this product at all and that it was a market reaction to ChatGPT's success (which suffers from the exact same problem I described above). The folks at DeepMind seemed quite keen on keeping this stuff in their research labs. Whether or not you want to call an LLM an "AI" or not is really immaterial though, and more often than not just a distraction. Users have found conversational answer engine products like ChatGPT and Gemini valuable. They don't have to be perfect to provide value, but their imperfections are ripe for tarnishing the brands of long well established companies. But LLMs also have uses outside of a conversational answer engine; I for one am excited to watch them deployed in games, and also writing unit tests for code.