Topics
later
AI
Amazon
Image Credits:efenzi / Getty Images
Apps
Biotech & Health
clime
Image Credits:efenzi / Getty Images
Cloud Computing
Commerce
Crypto
Enterprise
EVs
Fintech
Fundraising
Gadgets
game
Government & Policy
computer hardware
layoff
Media & Entertainment
Meta
Microsoft
secrecy
Robotics
Security
Social
infinite
Startups
TikTok
Transportation
Venture
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
picture
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
“ Running with scissors hold is a cardio exercise that can increase your heart rate and require absorption and focus , ” says Google ’s newAI hunt feature . “ Some say it can also improve your pores and give you forte . ”
Google ’s AI feature pulled this reply from a website calledLittle Old Lady Comedy , which , as its name makes clear , is a funniness blog . But the gaffe is so silly that it ’s been circulating on societal medium , along with other obviously incorrect AI overviews on Google . efficaciously , casual users are now red teaming these product on societal medium .
In cybersecurity , some companies will hire “ red teams ” – honorable hacker – who attempt to breach their products as though they ’re bad actors . If a flushed team get hold a vulnerability , then the troupe can desexualize it before the intersection ships . Google certainly conducted a form of red teaming before releasing an AI product on Google Search , which isestimatedto process one million million of queries per day .
It ’s surprising , then , when a highly resourced company like Google still ship product with obvious fault . That ’s why it ’s now become a meme to clown around on the failure of AI products , particularly in a clock time when AI is becoming more ubiquitous . We ’ve seen this with spoiled spelling onChatGPT , video source ’ failure to understandhow humans corrode spaghetti , andGrok AInews summary on X that , like Google , do n’t interpret satire . But these meme could really serve as useful feedback for companies develop and testing AI .
Despite the eminent - profile nature of these flaw , tech companies often minimise their shock .
“ The exemplar we ’ve seen are generally very uncommon query , and are n’t representative of most people ’s experiences , ” Google tell TechCrunch in an emailed statement . “ We impart panoptic examination before found this new experience , and will utilise these isolated examples as we continue to refine our system overall . ”
Not all user see the same AI termination , and by the time a particularly bad AI mesmerism stimulate around , the military issue has often already been reclaim . In a more late event that went viral , Google indicate that if you ’re making pizza pie butthe cheese wo n’t stick , you could add about an eighth of a cup of gum to the sauce to “ give it more tackiness . ” As it wrick out , the AI is pull this response froman eleven - year - old Reddit commentfrom a user named “ f––smith . ”
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
Google AI overview intimate bring glue to get cheese to stick to pizza , and it turns out the source is an 11 year old Reddit comment from user F*cksmith 😂 pic.twitter.com/uDPAbsAKeO
Beyond being an incredible blunder , it also indicate that AI cognitive content deals may be overvalued . Google has a$60 million contractwith Reddit to license its content for AI model training , for instance . Reddit signed a interchangeable deal withOpenAIlast week , and Automattic properties WordPress.org and Tumblr arerumoredto be in talk to sell data point to Midjourney and OpenAI .
To Google ’s credit , a lot of the errors that are circulating on societal media come from unconventional searches design to trip up up the AI . At least I hope no one is seriously searching for “ health benefits of running with scissor grip . ” But some of these screw - ups are more serious . scientific discipline diarist Erin Rossposted on Xthat Google sprinkle out incorrect data about what to do if you get a rattlesnake raciness .
Ross ’s post , which get over 13,000 likes , shows that AI recommended apply a tourniquet to the wounding , cutting the combat injury and suck in out the spitefulness . According to theU.S. Forest Service , these are all things you shouldnotdo , should you get bitten . Meanwhile on Bluesky , the author T Kingfisher hyperbolise a military post that shows Google ’s Geminimisidentifying a poisonous mushroomas a mutual bloodless button mushroom – screenshots of the billet havespreadto other program as a cautionary tale .
Good ol’ Google AI : telling you to do the accurate things you * are not speculate to do * when bitten by a rattler . From mushroom to snakebites , AI content is really dangerous.pic.twitter.com/UZXgBjsre9
When a bad AI reply function viral , the AI could get more baffled by the newfangled capacity around the topic that comes about as a result . On Wednesday , New York Times reporter Aric Toler posteda screenshot on Xthat shows a question need if a dog has ever played in the NHL . The AI ’s response was yes – for some reason , the AI called the Calgary Flames actor Martin Pospisil a dog . Now , when you make that same interrogation , the AI pulls up an clause fromthe Daily Dotabout how Google ’s AI continue cogitate that dogs are playing sportsman . The AI is being give its own mistakes , poison it further .
This is the implicit in problem of training these large - scale AI model on the internet : sometimes , the great unwashed on the internet prevarication . But just like how there’sno rule against a domestic dog playing basketball game , there ’s unluckily no regulation against big tech companies ship bad AI products .
As the locution goes : garbage in , refuse out .
