Say you want to use Gemini to create a marketing campaign, and you ask it to generate 10 pictures of \u201ca person walking a dog in a park.\u201d Because you don\u2019t specify the type of person, dog, or park, it\u2019s dealer\u2019s choice \u2014 the generative model will put out what it is most familiar with. And in many cases, that is a product not of reality, but of the training data, which can have all kinds of biases baked in.<\/p>\n
What kinds of people, and for that matter dogs and parks, are most common in the thousands of relevant images the model has ingested? The fact is that white people are over-represented in a lot of these image collections (stock imagery, rights-free photography, etc.), and as a result the model will default to white people in a lot of cases if you don\u2019t specify.<\/p>\n
That\u2019s just an artifact of the training data, but as Google points out, \u201cbecause our users come from all over the world, we want it to work well for everyone. If you ask for a picture of football players, or someone walking a dog, you may want to receive a range of people. You probably don\u2019t just want to only receive images of people of just one type of ethnicity (or any other characteristic).\u201d<\/p>\n
![\"Illustration](\"https:\/\/techcrunch.com\/wp-content\/uploads\/2023\/10\/GettyImages-1421655800.jpg\")
<\/p>\n
Imagine asking for an image like this \u2014 what if it was all one type of person? Bad outcome! Image Credits:<\/strong> Getty Images \/ victorikart<\/p>\n<\/div>\nNothing wrong with getting a picture of a white guy walking a golden retriever in a suburban park. But if you ask for 10, and they\u2019re all<\/em> white guys walking goldens in suburban parks? And you live in Morocco, where the people, dogs, and parks all look different? That\u2019s simply not a desirable outcome. If someone doesn\u2019t specify a characteristic, the model should opt for variety, not homogeneity, despite how its training data might bias it.<\/p>\nThis is a common problem across all kinds of generative media. And there\u2019s no simple solution. But in cases that are especially common, sensitive, or both, companies like Google, OpenAI, Anthropic, and so on invisibly include extra instructions for the model.<\/p>\n
I can\u2019t stress enough how commonplace this kind of implicit instruction is. The entire LLM ecosystem is built on implicit instructions \u2014 system prompts, as they are sometimes called, where things like \u201cbe concise,\u201d \u201cdon\u2019t swear,\u201d and other guidelines are given to the model before every conversation. When you ask for a joke, you don\u2019t get a racist joke \u2014 because despite the model having ingested thousands of them, it has also been trained, like most of us, not to tell those. This isn\u2019t a secret agenda (though it could do with more transparency), it\u2019s infrastructure.<\/p>\n
Where Google\u2019s model went wrong was that it failed to have implicit instructions for situations where historical context was important. So while a prompt like \u201ca person walking a dog in a park\u201d is improved by the silent addition of \u201cthe person is of a random gender and ethnicity\u201d or whatever they put, \u201cthe U.S. Founding Fathers signing the Constitution\u201d is definitely not improved by the same.<\/p>\n
As the Google SVP Prabhakar Raghavan put it:<\/p>\n
\nFirst, our tuning to ensure that Gemini showed a range of people failed to account for cases that should clearly not show a range. And second, over time, the model became way more cautious than we intended and refused to answer certain prompts entirely \u2014 wrongly interpreting some very anodyne prompts as sensitive.<\/p>\n
These two things led the model to overcompensate in some cases, and be over-conservative in others, leading to images that were embarrassing and wrong.<\/p>\n<\/blockquote>\n
I know how hard it is to say \u201csorry\u201d sometimes, so I forgive Raghavan for stopping just short of it. More important is some interesting language in there: \u201cThe model became way more cautious than we intended.\u201d<\/p>\n
Now, how would a model \u201cbecome\u201d anything? It\u2019s software. Someone \u2014 Google engineers in their thousands \u2014 built it, tested it, iterated on it. Someone wrote the implicit instructions that improved some answers and caused others to fail hilariously. When this one failed, if someone could have inspected the full prompt, they likely would have found the thing Google\u2019s team did wrong.<\/p>\n
Google blames the model for \u201cbecoming\u201d something it wasn\u2019t \u201cintended\u201d to be. But they made the model! It\u2019s like they broke a glass, and rather than saying \u201cwe dropped it,\u201d they say \u201cit fell.\u201d (I\u2019ve done this.)<\/p>\n
Mistakes by these models are inevitable, certainly. They hallucinate, they reflect biases, they behave in unexpected ways. But the responsibility for those mistakes does not belong to the models \u2014 it belongs to the people who made them. Today that\u2019s Google. Tomorrow it\u2019ll be OpenAI. The next day, and probably for a few months straight, it\u2019ll be X.AI.<\/p>\n
These companies have a strong interest in convincing you that AI is making its own mistakes. Don\u2019t let them.<\/p>\n<\/p><\/div>\n
\n
Source link <\/a><\/p>\n","protected":false},"excerpt":{"rendered":"Google has apologized (or come very close to apologizing) for another embarrassing AI blunder this week, an image-generating model that injected diversity into pictures with a farcical disregard for historical context. While the underlying issue is perfectly understandable, Google blames the model for \u201cbecoming\u201d oversensitive. But the model didn\u2019t make itself, guys. The AI system […]<\/p>\n","protected":false},"author":1,"featured_media":77970,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[14],"tags":[],"class_list":{"0":"post-77969","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-tech"},"_links":{"self":[{"href":"https:\/\/entertainment.runfyers.com\/index.php\/wp-json\/wp\/v2\/posts\/77969","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/entertainment.runfyers.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/entertainment.runfyers.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/entertainment.runfyers.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/entertainment.runfyers.com\/index.php\/wp-json\/wp\/v2\/comments?post=77969"}],"version-history":[{"count":0,"href":"https:\/\/entertainment.runfyers.com\/index.php\/wp-json\/wp\/v2\/posts\/77969\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/entertainment.runfyers.com\/index.php\/wp-json\/wp\/v2\/media\/77970"}],"wp:attachment":[{"href":"https:\/\/entertainment.runfyers.com\/index.php\/wp-json\/wp\/v2\/media?parent=77969"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/entertainment.runfyers.com\/index.php\/wp-json\/wp\/v2\/categories?post=77969"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/entertainment.runfyers.com\/index.php\/wp-json\/wp\/v2\/tags?post=77969"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}