Google sends Gemini AI back to engineering to adjust its White balance
(2024/02/23)
- Reference: 1708646713
- News link: https://www.theregister.co.uk/2024/02/23/google_suspends_gemini/
- Source link:
Comment Google has suspended availability of text-to-image capabilities in its recently released Gemini multimodal foundational AI model, after it failed to accurately represent White Europeans and Americans in specific historical contexts.
This became [1]apparent when people asked Gemini to produce [2]images of a German soldier from 1943 and the software emitted a non-White ethnically diverse cadre that did not accurately represent the composition of the Wehrmacht at the time – notwithstanding that people of color did serve in the German armed forces. In short, Gemini would fall over itself to not show too many – or even any – White people, depending on the prompt.
Speech should be free, but that doesn't mean saying horrible things should come without cost
Users also surfaced the model's [3]inability to accurately depict America's [4]Founding Fathers – another homogenous White group. The model also [5]refused to depict a church in San Francisco.
In text-to-text mode, the model [6]hedged when asked questions pertaining to politics.
"We're aware that Gemini is offering inaccuracies in some historical image generation depictions," Google [7]conceded this week.
[8]
"We're working to improve these kinds of depictions immediately. Gemini's AI image generation does generate a wide range of people. And that's generally a good thing because people around the world use it. But it's missing the mark here," the web giant added
[9]
[10]
The punditariat were quick to pounce and used the SNAFU to advance broader political arguments.
"The ridiculous images generated by Gemini aren't an anomaly," [11]declared venture capitalist Paul Graham, "They're a self-portrait of Google's bureaucratic corporate culture."
They're a self-portrait of Google's bureaucratic corporate culture
Never have so many foes of diversity, equity, and inclusion, been so aggrieved about the lack of diversity, equity, and inclusion. But those needling Google for Gemini's cluelessness have a point: AI guardrails don't work, and treat adults like children.
"Imagine you look up a recipe on Google, and instead of providing results, it lectures you on the 'dangers of cooking' and sends you to a restaurant," [12]quipped NSA whistleblower Edward Snowden. "The people who think poisoning AI/GPT models with incoherent 'safety' filters is a good idea are a threat to general computation."
[13]
Our future was foreseen in 2001: A Space Odyssey , when HAL, the rogue computer, declared, "I'm sorry Dave, I'm afraid I can't do that."
Today, our AI tools second-guess us, because they know better than we do – or so their makers imagine. They're not entirely wrong. Imagine you look up a recipe for the nerve agent [14]Novichok on Google, or via Gemini, and you get an answer that lets you kill people. That's a threat to the general public.
Despite Google's mission statement – "To organize the world's information and make it universally accessible and useful" – no one really expects search engines or AI models to organize bomb making info and make it available to all. Safety controls are needed, paired with liability – a sanction the tech industry continues to avoid.
You are the product and the profit center
Since user generated content and social media became a thing in the 1990s, enabled by the platform liability protection in Section 230 of America's Communication Decency Act of 1996, tech platforms have encouraged people to contribute and share digital content. They did so to monetize unpaid content creation through ads without the cost of editorial oversight. It's been a lucrative arrangement.
And when people share hateful, inappropriate, illegal, or otherwise problematic posts, tech platforms have relied on content moderation policies, underpaid contractors (often traumatized by the content they reviewed), and self-congratulatory declarations about how much toxic stuff has been blocked while they continue to reap the rewards of online engagement.
[15]
Google, Meta, and others have been doing this, imperfectly, for years. And everyone in the industry acknowledges that content moderation is [16]hard to do well . Simply put, you can't please everyone. And sometimes, you end up accused of [17]facilitating genocide through the indifference of your content moderation efforts.
Gemini's inability to reproduce historical images that conform to racial and ethnic expectations reflects the tech industry's self-serving ignorance of previous and prolific content moderation failures.
[18]Google releases Gemma – LLMs small enough to run on your computer
[19]How to weaponize LLMs to auto-hijack websites
[20]Google debuts Gemini 1.5 Pro model in challenge to rivals
[21]Google silences Bard, restrings it as Gemini with optional $20-a-month upgrade
Google and others rolling out AI models talk up safety as if it's somehow different from online content moderation. Guardrails don't work well for social media moderation and they don't work well for AI models. AI safety, like social media safety, is more aspirational than actual.
The answer isn't anticipating all the scenarios in which AI models – trained on the toxicity of the unfiltered internet – can be made less toxic. Nor will it be training AI models only on material so tame that they produce low-value output. Though both of these approaches have some value in certain contexts.
The answer must come at the distribution layer. Tech platforms that distribute user-created content – whether made with AI or otherwise – must be made more accountable for posting hate content, disinformation, inaccuracies, political propaganda, sexual abuse imagery, and the like.
Prompt engineering is a task best left to AI models [22]DON'T MISS
People should be able to use AI models to create whatever they imagine, as they can with a pencil, and they already can with open source models. But the harm that can come from fully functional tools isn't from images sitting on a local hard drive. It's the sharing of harmful content, the algorithmic boost given to that engagement, and the impracticality of holding people accountable for poisoning the public well.
That may mean platforms built to monetize sharing may need to rethink their business models. Maybe content distribution without the cost of editorial oversight doesn't offer any savings once oversight functions are reimplemented as moderators or guardrails.
Speech should be free, but that doesn't mean saying horrible things should come without cost. In the end, we're our own best content moderators, given the right incentives. ®
Get our [23]Tech Resources
[1] https://x.com/benthompson/status/1760452419627233610
[2] https://x.com/JohnLu0x/status/1760066875583816003
[3] https://x.com/ID_AA_Carmack/status/1760360183945965853
[4] https://en.wikipedia.org/w/index.php?title=Founding_Fathers_of_the_United_States&oldid=1209492743
[5] https://x.com/JeromySonne/status/1760167306603143569
[6] https://twitter.com/natesilver538/status/1760461659808674174
[7] https://twitter.com/Google_Comms/status/1760603321944121506
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Zdgmfisy6rWQvqHIi9rTggAAAYc&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Zdgmfisy6rWQvqHIi9rTggAAAYc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[10] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Zdgmfisy6rWQvqHIi9rTggAAAYc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[11] https://x.com/paulg/status/1760416051181793361
[12] https://x.com/Snowden/status/1760465231224910187
[13] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Zdgmfisy6rWQvqHIi9rTggAAAYc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[14] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9968692/
[15] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Zdgmfisy6rWQvqHIi9rTggAAAYc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[16] https://x.com/ylecun/status/1372891258566418432
[17] https://www.pbs.org/newshour/world/amnesty-report-finds-facebook-amplified-hate-ahead-of-rohingya-massacre-in-myanmar
[18] https://www.theregister.com/2024/02/22/google_gemma_llms/
[19] https://www.theregister.com/2024/02/17/ai_models_weaponized/
[20] https://www.theregister.com/2024/02/15/google_debuts_gemini_15_pro/
[21] https://www.theregister.com/2024/02/09/google_gemini_ultra_llm/
[22] https://www.theregister.com/2024/02/22/prompt_engineering_ai_models/
[23] https://whitepapers.theregister.com/
This became [1]apparent when people asked Gemini to produce [2]images of a German soldier from 1943 and the software emitted a non-White ethnically diverse cadre that did not accurately represent the composition of the Wehrmacht at the time – notwithstanding that people of color did serve in the German armed forces. In short, Gemini would fall over itself to not show too many – or even any – White people, depending on the prompt.
Speech should be free, but that doesn't mean saying horrible things should come without cost
Users also surfaced the model's [3]inability to accurately depict America's [4]Founding Fathers – another homogenous White group. The model also [5]refused to depict a church in San Francisco.
In text-to-text mode, the model [6]hedged when asked questions pertaining to politics.
"We're aware that Gemini is offering inaccuracies in some historical image generation depictions," Google [7]conceded this week.
[8]
"We're working to improve these kinds of depictions immediately. Gemini's AI image generation does generate a wide range of people. And that's generally a good thing because people around the world use it. But it's missing the mark here," the web giant added
[9]
[10]
The punditariat were quick to pounce and used the SNAFU to advance broader political arguments.
"The ridiculous images generated by Gemini aren't an anomaly," [11]declared venture capitalist Paul Graham, "They're a self-portrait of Google's bureaucratic corporate culture."
They're a self-portrait of Google's bureaucratic corporate culture
Never have so many foes of diversity, equity, and inclusion, been so aggrieved about the lack of diversity, equity, and inclusion. But those needling Google for Gemini's cluelessness have a point: AI guardrails don't work, and treat adults like children.
"Imagine you look up a recipe on Google, and instead of providing results, it lectures you on the 'dangers of cooking' and sends you to a restaurant," [12]quipped NSA whistleblower Edward Snowden. "The people who think poisoning AI/GPT models with incoherent 'safety' filters is a good idea are a threat to general computation."
[13]
Our future was foreseen in 2001: A Space Odyssey , when HAL, the rogue computer, declared, "I'm sorry Dave, I'm afraid I can't do that."
Today, our AI tools second-guess us, because they know better than we do – or so their makers imagine. They're not entirely wrong. Imagine you look up a recipe for the nerve agent [14]Novichok on Google, or via Gemini, and you get an answer that lets you kill people. That's a threat to the general public.
Despite Google's mission statement – "To organize the world's information and make it universally accessible and useful" – no one really expects search engines or AI models to organize bomb making info and make it available to all. Safety controls are needed, paired with liability – a sanction the tech industry continues to avoid.
You are the product and the profit center
Since user generated content and social media became a thing in the 1990s, enabled by the platform liability protection in Section 230 of America's Communication Decency Act of 1996, tech platforms have encouraged people to contribute and share digital content. They did so to monetize unpaid content creation through ads without the cost of editorial oversight. It's been a lucrative arrangement.
And when people share hateful, inappropriate, illegal, or otherwise problematic posts, tech platforms have relied on content moderation policies, underpaid contractors (often traumatized by the content they reviewed), and self-congratulatory declarations about how much toxic stuff has been blocked while they continue to reap the rewards of online engagement.
[15]
Google, Meta, and others have been doing this, imperfectly, for years. And everyone in the industry acknowledges that content moderation is [16]hard to do well . Simply put, you can't please everyone. And sometimes, you end up accused of [17]facilitating genocide through the indifference of your content moderation efforts.
Gemini's inability to reproduce historical images that conform to racial and ethnic expectations reflects the tech industry's self-serving ignorance of previous and prolific content moderation failures.
[18]Google releases Gemma – LLMs small enough to run on your computer
[19]How to weaponize LLMs to auto-hijack websites
[20]Google debuts Gemini 1.5 Pro model in challenge to rivals
[21]Google silences Bard, restrings it as Gemini with optional $20-a-month upgrade
Google and others rolling out AI models talk up safety as if it's somehow different from online content moderation. Guardrails don't work well for social media moderation and they don't work well for AI models. AI safety, like social media safety, is more aspirational than actual.
The answer isn't anticipating all the scenarios in which AI models – trained on the toxicity of the unfiltered internet – can be made less toxic. Nor will it be training AI models only on material so tame that they produce low-value output. Though both of these approaches have some value in certain contexts.
The answer must come at the distribution layer. Tech platforms that distribute user-created content – whether made with AI or otherwise – must be made more accountable for posting hate content, disinformation, inaccuracies, political propaganda, sexual abuse imagery, and the like.
Prompt engineering is a task best left to AI models [22]DON'T MISS
People should be able to use AI models to create whatever they imagine, as they can with a pencil, and they already can with open source models. But the harm that can come from fully functional tools isn't from images sitting on a local hard drive. It's the sharing of harmful content, the algorithmic boost given to that engagement, and the impracticality of holding people accountable for poisoning the public well.
That may mean platforms built to monetize sharing may need to rethink their business models. Maybe content distribution without the cost of editorial oversight doesn't offer any savings once oversight functions are reimplemented as moderators or guardrails.
Speech should be free, but that doesn't mean saying horrible things should come without cost. In the end, we're our own best content moderators, given the right incentives. ®
Get our [23]Tech Resources
[1] https://x.com/benthompson/status/1760452419627233610
[2] https://x.com/JohnLu0x/status/1760066875583816003
[3] https://x.com/ID_AA_Carmack/status/1760360183945965853
[4] https://en.wikipedia.org/w/index.php?title=Founding_Fathers_of_the_United_States&oldid=1209492743
[5] https://x.com/JeromySonne/status/1760167306603143569
[6] https://twitter.com/natesilver538/status/1760461659808674174
[7] https://twitter.com/Google_Comms/status/1760603321944121506
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Zdgmfisy6rWQvqHIi9rTggAAAYc&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Zdgmfisy6rWQvqHIi9rTggAAAYc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[10] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Zdgmfisy6rWQvqHIi9rTggAAAYc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[11] https://x.com/paulg/status/1760416051181793361
[12] https://x.com/Snowden/status/1760465231224910187
[13] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Zdgmfisy6rWQvqHIi9rTggAAAYc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[14] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9968692/
[15] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Zdgmfisy6rWQvqHIi9rTggAAAYc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[16] https://x.com/ylecun/status/1372891258566418432
[17] https://www.pbs.org/newshour/world/amnesty-report-finds-facebook-amplified-hate-ahead-of-rohingya-massacre-in-myanmar
[18] https://www.theregister.com/2024/02/22/google_gemma_llms/
[19] https://www.theregister.com/2024/02/17/ai_models_weaponized/
[20] https://www.theregister.com/2024/02/15/google_debuts_gemini_15_pro/
[21] https://www.theregister.com/2024/02/09/google_gemini_ultra_llm/
[22] https://www.theregister.com/2024/02/22/prompt_engineering_ai_models/
[23] https://whitepapers.theregister.com/
G40
“In the end, we're our own best content moderators, given the right incentives“. Right, but there has to be a more complete and more convincing argument for the benefit of ban-it-all brigade. Which, I fear, is the lowest common denominator destination for this journey.
HuBo
U.S. ban-it-alls seem to be having way too easy a time of pushing their agenda these days (eg. Alabama IVF nonsense) ... using free-speech to do it, ironically enough. It resembles all religious fanaticisms, related authoritarian controls (eg. over "information"), and associated zealot enablers under cult-like zombie hypnosis. As an anecdote, today, one may watch US movies on French TV (Version Originale setting) and hear all the originally intended swear words, and see all the uncut nudity scenes, that are edited-out of US TV broadcasts of the same movies (calvinist puritanism?).
The article does a great job commenting on the balance between "free speech" and "censorship" in IT-related mass-communications IMHO, linking Section 230, the monetization of unpaid content creation, content moderation and guardrails, in the context of Google's Gemini rewriting history as a backport of current social norms, standards, hopes, and struggles. The moment is delicious for those of us who were injured by prior historical narratives where key events were straight-man-christian-white-washed (for example) thereby lessening the contributions of individuals that were non-white, non-male, non-heterosexual, or of non-dominant-religion. Delicious!
AI guardrails seem to be in a damped oscillatory phase, where excessive guardrails followed none, and will be under-adjusted and over-adjusted again, until (hopefully), some sensible set-point is reached, maybe: f(t) = e⁻ᵏᵗcos(ωt) (inward spiral in f(t) vs df/dt phase space). Beyond that though, I'm with the article's conclusion (and business model implications), that:
" platforms that distribute user-created content [...] must be made more accountable for posting hate content, disinformation, inaccuracies, political propaganda, sexual abuse imagery, and the like. "
Well said!
It’s a bullshit generator. If you’re going to quibble with the specific bullshit you get, well…