Topics
Latest
AI
Amazon
Image Credits:Getty Images (Image has been modified)
Apps
Biotech & Health
Climate
Image Credits:Getty Images (Image has been modified)
Cloud Computing
Commerce
Crypto
Image Credits:Anthropic
initiative
EVs
Fintech
Image Credits:Future of Life Institute
fundraise
contraption
Gaming
Government & Policy
Hardware
Layoffs
Media & Entertainment
Meta
Microsoft
privateness
Robotics
surety
Social
Space
inauguration
TikTok
Transportation
speculation
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
Last week , OpenAI launchedAdvanced Voice Mode with Vision , which feeds substantial - meter telecasting toChatGPT , grant the chatbot to “ see ” beyond the confines of its app bed . The premise is that by giving ChatGPT greater contextual awareness , that bot can answer in a more natural and intuitive way .
But the first time I hear it , it lied to me .
“ That sofa see comfortable ! ” ChatGPT said as I deem up my phone and asked the bot to describe our animation way . It had mistaken the ottoman for a couch .
“ My fault ! ” ChatGPT said when I correct it . “ Well , it still looks like a comfortable space . ”
It ’s been intimately a year since OpenAI firstdemoedAdvanced Voice Mode with Vision , which the companypitchedas a gradation toward AI as depicted in the Spike Jonze movie “ Her . ” The mode OpenAI sold it , Advanced Voice Mode with Vision would grant ChatGPT superpowers — enabling the bot to resolve sketched - out mathematics problems , interpret emotion , and respond toaffectionate letters .
Has it achieve all that ? More or less . But Advanced Voice Mode with Vision has n’t solved ChatGPT ’s biggest outcome : reliableness . If anything , the feature make the bot’shallucinationsmore obvious .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
At one point , curious to see if Advanced Voice Mode with Vision could assist ChatGPT offer fashion pointers , I enable it and need ChatGPT to rate an getup of mine . It jubilantly did so . But while the bot would give opinion on my jeans and olive - colored - shirt combo , it systematically missed the chocolate-brown jacket I was wear out .
I ’m not the only one who has happen slipups .
When OpenAI prexy Greg Brockman show off Advanced Voice Mode with Vision on “ 60 Minutes ” earlier this month , ChatGPT made a misunderstanding on a geometry problem . When calculating the area of a Triangulum , itmisidentifiedthe triangle ’s altitude .
So my question is , what good is “ Her”-like AI if you ca n’t trust it ?
With each ChatGPT misfire , I feel myself becoming less and less inclined to reach into my pocket , unlock my speech sound , launch ChatGPT , assailable Advanced Voice Mode , and enable Vision — a cumbersome series of tone in the good of circumstance . With its lustrous and gay behaviour , Advanced Voice Mode is clearly design to engender trust . When it does n’t deliver on that inexplicit promise , it ’s jarring — and disappointing .
Perhaps OpenAI can work out the hallucination problem once and for all someday . Until then , we ’re stuck with a bot that views the world through criss - crossed wiring . And honestly , I ’m not sure who might want that .
News
OpenAI ’s 12 days of “ shipmas ” continues : OpenAI is releasing new products every day up until December 20.Here’sa roundup of all the announcements , which we ’re update regularly .
YouTube let creators opt out : YouTube is give Maker more selection over how third party can utilize their content to prepare their AI model . Lord and right holders will be able to flag for YouTube if they ’re countenance specific company to train models on their clips .
Meta ’s smart glasses get rise : Meta ’s Ray - Ban Meta smart glasses have gottenseveral new AI - powered update , include the power to have an ongoing conversation with Meta ’s AI and read between spoken language .
DeepMind ’s answer to Sora : Google DeepMind , Google ’s flagship AI research lab , want to nonplus OpenAI at the video - generation game . On Monday , DeepMind announce Veo 2 , a next - gen video recording - generating AI that can create two - minute - plus clips in resolution up to 4k ( 4,096 x 2,160 pel ) .
OpenAI whistle-blower found bushed : A former OpenAI employee , Suchir Balaji , was recently find utter in his San Francisco flat , according to the San Francisco Office of the Chief Medical Examiner . In October , the 26 - year - old AI researcher raised concern about OpenAI bring out right of first publication law when he was interviewed by The New York Times .
Grammarly acquires Coda : Grammarly , best known for its elan and spell - check tools , has acquired productiveness inauguration Coda for an unrevealed amount . As part of the deal , Coda ’s CEO and co - founder , Shishir Mehrotra , will become the young CEO of Grammarly .
Cohere is working with Palantir : TechCrunch exclusively reported that Cohere , the enterprise - focus AI startup assess at $ 5.5 billion , has a partnership with information analytics house Palantir . Palantir is vocal about its close — and at timescontroversial — work with U.S. defense and intelligence service authority .
Research paper of the week
Anthropic has pulled back the pall on Clio ( “ Claudeinsights andobservations ” ) , a organization that the company uses to understand how customers are employing its various AI models . Clio , which Anthropic compare to analytics tools such as Google Trends , is ply “ worthful insights ” for improving the safety of Anthropic ’s AI , take the company .
Anthropic tapped Clio to compose anonymized usage datum , some of which the company made public last week . So what are customers using Anthropic ’s AI for ? A range of a function of tasks — but web and nomadic app developing , capacity innovation , and donnish research top the list . Predictably , the use cases vary across spoken language ; for example , Japanese speakers are more probable to ask Anthropic ’s AI to analyse gum anime than Spanish utterer .
Model of the week
AI startupPikareleased its next - gen television generation model , Pika 2 , which can create a clip from a lineament , target , and location that users supply . Via Pika ’s program , users can upload multiple references ( for example , simulacrum of a council chamber and office doer ) and Pika 2 will “ intuit ” the use of each reference before combining them into a single scene .
Now , no manakin ’s perfect , of course . See the “ Zanzibar copal ” below created by Pika 2 , which has telling consistency but suffers from the aesthetic bizarreness present in all reproductive AI footage .
pic.twitter.com/3jWCy4659oLike I pronounce , Animes will be the first musical genre that s 100 % AI father . Its amazing to see what ’s already possible with Pika 2.0
— Chubby ♨ ️ ( @kimmonismus)December 16 , 2024
Still , the tools are very chop-chop improve in the video domain — and in equal component pique the interest and raising the anger of creatives .
Grab bag
The Future of Life Institute ( FLI ) , the nonprofit organization co - found by MIT cosmologist Max Tegmark , released an “ AI Safety Index ” designed to evaluate the base hit drill of leading AI companies across five fundamental areas : current hurt , safety framework , experiential base hit scheme , governance and accountability , and transparency and communication .
Meta was the bad of the cluster evaluated on the Index , with an overall F grade . ( The Index apply a numerical and GPA - based scoring system . ) Anthropic was the best but failed to manage better than a C — suggesting that there ’s room for improvement .
