Topics
Latest
AI
Amazon
Image Credits:Joshua Lott / Getty Images
Apps
Biotech & Health
clime
Image Credits:Joshua Lott / Getty Images
Cloud Computing
Department of Commerce
Crypto
Image Credits:Apple
enterprisingness
EVs
Fintech
Image Credits:MIT CSAIL
Fundraising
Gadgets
game
Image Credits:kudzueye(opens in a new window)
Government & Policy
Hardware
layoff
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
surety
societal
Space
Startups
TikTok
Transportation
Venture
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
meet Us
Hiya , folks , welcome to TechCrunch ’s even AI newssheet . If you want this in your inbox every Wednesday , signal uphere .
The billionaires are fighting again .
On Monday , Elon Musk , the world ’s richest humans , offered to buy the nonprofitthat effectively governs OpenAI for $ 97.4 billion . In reply to Musk ’s offer , OpenAI CEO Sam Altman before Monday authoreda brash station on X , writing , “ No give thanks you , but we will bribe Twitter for $ 9.74 billion if you need . ” ( Musk and investors famouslypurchased Twitter for $ 44 billionin 2022 . )
Musk ’s dictation , serious or not , may complicate OpenAI ’s effort to convert to a for - profit public welfare corporation within two class . Now OpenAI ’s board will have to march it ’s not underselling OpenAI ’s nonprofit by handing the nonprofit ’s assets , including IP from OpenAI ’s research , to an insider ( e.g. , Altman ) for a discount .
OpenAI could make the case that Musk ’s bidding is a hostile takeover attempt given that Musk and Altmanaren’t the best of friends . It could also argue that Musk ’s offer is n’t believable because OpenAI is already in the thick of a restructuring outgrowth . Or OpenAI couldchallenge Musk on whether he has the funds .
In astatement Tuesday , Andy Nussbaum , outside counselor representing OpenAI ’s board , said that Musk ’s bid “ does n’t set a time value for [ OpenAI ’s ] nonprofit ” and that the non-profit-making is “ not for cut-rate sale . ” Nussbaum added , “ Respectfully , it is not up to a challenger to resolve what is in the best interests of OpenAI ’s mission . ”My colleague Maxwell Zeff and Iwrote a more detailed pieceon what to gestate in the coming weeks . But guaranteed , Musk ’s offer — not to mention hisongoing causa against OpenAI over what he lay claim is deceitful conduct — promises to make for fierce court do .
News
Apple ’s new robot : Apple create a enquiry robot that takes a page from Pixar ’s playbook . The company ’s robotic lamp run as a more energising version of a HomePod or other smart speaker . The person facing the lamp asks a enquiry , and the robot answer in Siri ’s phonation .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
Is AI get us dumb?:Researchers of late publish a survey looking at how using productive AI at work affects critical thinking skills . It see that when we bank too much on AI to recollect for us , we get worse at solving problems ourselves when AI fail .
AI for all , perhaps : In a novel essay on his personal blog , Altman admitted that AI ’s welfare may not be widely distributed — and said that OpenAI is open to “ strange - sounding ” ideas like a “ compute budget ” to “ enable everyone on Earth to use a lot of AI . ”
Christie ’s disceptation : Fine artistry auction sale house Christie ’s has sold AI - generated art before . But soon it plans to control its first show dedicated solely to works created with AI , an announcement that has been meet with assorted reviews — and a petition call for the vendue ’s cancellation .
Better than gold : An AI system develop by Google DeepMind , Google ’s conduct AI research lab , appears to have surpassed the average gold medalist in lick geometry problems in an international mathematics competitor .
Research paper of the week
We cognize that most AI models ca n’t do canonical tasks dependably , like solving grade - school - level math problems . What we do n’t always know isthe reasonbehind their failures . According to a team of researchers at MIT CSAIL , erroneous bench mark may be in part to fault .
In a new sketch , the MIT CSAIL researchers ascertain that while today ’s top - performing models still make real mistakes on democratic AI bench mark , over 50 % of “ model error ” are really because of mislabeled and ambiguous questions in those benchmarks .
“ If we want to properly measure modeling reliableness , we need to rethink how we manufacture benchmarks to minimize label errors , ” saidone of the investigator , MIT faculty phallus and OpenAI staffer Aleksander Madry , in a post on X. “ This is just a first tone . ”
Model of the week
You ’ve get a line of deepfakes before . But what about deepfakes of boring everyday scenes ? That ’s the mind behindBoring Reality Hunyuan LoRA ( Boreal - HL ) , a finely - tuned AI video source that excels at creating videos of … well , jolly commonplace stuff and nonsense .
Boreal - HL can get clip of holidaymaker eating deoxyephedrine cream , people barbecuing inwardness , people in luncheon meetings , executives giving talking to at conference , couples at marriage ceremony , and other routine slices of life . This newsman finds the absurdity of the thing screaming — peculiarly considering how Laputan it is to run . It take Boreal - HL at least five minutes to generate a single magazine .
Grab bag
Thanks to recent breakthroughs in AI efficiency , it ’s getting tacky — and easier — to coach extremely sophisticated models .
In a new composition , researchers at Shanghai Jiao Tong University and an AI company call up SII show that a example prepare on just 817 “ curated training samples ” can outperform models trained on 100x more datum . The team exact that their model was even able to do certain doubt it had n’t seen during the preparation process , showing what they call “ out of arena ” capabilities .
The study follows on the heels of aStanford - conduce projectthat found it ’s possible to make an “ undefendable ” model rival OpenAI ’s o1 “ reasoning ” model for under $ 50 .
