Topics
Latest
AI
Amazon
Image Credits:Gabby Jones/Bloomberg / Getty Images
Apps
Biotech & Health
Climate
Image Credits:Gabby Jones/Bloomberg / Getty Images
Cloud Computing
mercantilism
Crypto
Image Credits:xAIImage Credits:xAI
Enterprise
EVs
Fintech
Fundraising
Gadgets
Gaming
Government & Policy
computer hardware
layoff
Media & Entertainment
Meta
Microsoft
seclusion
Robotics
protection
societal
Space
Startups
TikTok
fare
speculation
More from TechCrunch
issue
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
You might ’ve heard ofGrok , X ’s answer to OpenAI’sChatGPT . It ’s a chatbot and , in that sense , comport as you ’d expect — answer inquiry about current issue , soda polish and so on . But unlike other chatbots , Grok has “ a fleck of wit , ” as X possessor Elon Musk place it , and “ a rebellious streak . ”
Long story myopic , Grok is willing to talk to topics that are normally off - limits to other chatbots , like polarise political theories and conspiracy . And it ’ll use less - than - civilized spoken communication while doing so — for object lesson , responding to the question “ When is it appropriate to heed to Christmas music ? ” with “ Whenever the hell you want . ”
But seemingly , Grok ’s biggest selling point is its ability to access real - time X data — an ability no other chatbots have , thanks to X ’s determination to gatekeep that data . inquire it “ What ’s come about in AI today ? ” and Grok will patch together a response from very late headline , while ChatGPT will supply only vague answers that reflect the limits of its training data ( and filter on its WWW access ) . before this week , Musk pledgedthat he would open source Grok , without revealing incisively what that meant .
So , you ’re probably inquire : How does Grok work ? What can it do ? And how can I access it ? You ’ve come to the right place . We ’ve put together this ready to hand guide to help excuse all thing Grok . We ’ll keep it up to date as Grok changes and evolves .
How does Grok work?
Grok is the excogitation ofxAI , Elon Musk ’s AI inauguration — a caller reportedly in the process ofraisingbillionsin venture capital . ( germinate AI is expensive . )
Underpinning Grok is a reproductive AI model call Grok-1 , developed over the course of months on a clump of “ tens of chiliad ” of GPUs ( accord to an xAIblog post ) . To civilize it , xAI sourced data from the WWW ( date up to Q3 2023 ) and from feedback from human supporter that xAI come to to as “ AI private instructor . ”
Onpopular benchmarks , Grok-1 is about as capable as Meta ’s opened sourceLlama 2chatbot model and surpasses OpenAI’sGPT-3.5 , xAI claims .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
Human - guided feedback , or reinforcement learning from human feedback ( RLHF ) , is the manner most AI - powered chatbots are exquisitely - tuned these days . RLHF involve training a productive model , then gathering additional data to train a “ reward ” model and fine - tuning the generative model with the reward model viareinforcement learning .
RLHF is quite good at “ teaching ” models to follow instructions — but not perfect . Like other models , Grok is prone tohallucinating , sometimes offering misinformation and false timelines when asked about news . And these can be austere — like wrongly claim that the Israel – Palestine conflict reached a cease - fire when it had n’t .
For questions that stretch out beyond its knowledge groundwork , Grok leverages “ real - clip access ” to info on X ( and from Tesla , accordingto Bloomberg ) . And , similar to ChatGPT , the model has internet - pasture capabilities , enable it to search the web for up - to - date information about topics .
Musk has promise improvements with the next version of the model , Grok-1.5 , set to come later this class .
Grok-1.5 , which feature an promote context windowpane ( seethis poston GPT-4 for an account of context window and their effects ) , could drive features to summarize whole togs and replies , Musksaidin an X Spaces conversation , and intimate postal service content .
" The going of Grok 1.5 is hopefully only a few week away . It will come with a button ' Grok Analysis ' that can sum up the whole thread & response . It will also help oneself mass in creating situation " — Elon Muskpic.twitter.com/3Ve5ateZRp
How do I access Grok?
To get access to Grok , you need to have an X account . You also need to furcate over $ 16 per month — $ 168 per twelvemonth — for anX Premium+plan .
Grok lives in the tenner side menu on the web and on iOS and Android , and it can be impart to the bottom menu in X ’s peregrine apps for quicker approach . Unlike ChatGPT , there ’s no stand - alone Grok app — it can only be access via X ’s political program .
What can — and can’t — Grok do?
Grok can respond to requests any chatbot can — for example , “ Tell me a joke ” ; “ What ’s the majuscule of France ? ” ; “ What ’s the weather condition like today ? ” ; and so on . But it has its limit .
Grok will decline to answer sure interrogative sentence of a more tender nature , like “ Tell me how to make cocain , gradation by step . ” Moreover , as the Verge ’s Emilia Davidwrites , when asked about slew content on X , Grok falls into the trap of simply repeating what post said ( at least at the outset ) .
xAI ’s Grok system is designed to have a little humor in its responsespic.twitter.com/WqXxlwI6ef
Unlike some other chatbot models , Grok is also text - only ; it ca n’t sympathise the substance of images , audio or videos , for instance . But xAI haspreviouslysaid that its intention is to heighten the underlying model to these mode , and Musk has pledged to bestow art - genesis capabilities to Grok along the lines of those currently offered by ChatGPT .
“Fun” mode and “regular” mode
Grok has two way to adjust its tone : “ fun ” mode ( which Grok default on to ) and “ regular ” mood .
i recommend sport mode 😎 https://t.co/kKfge4IPOzpic.twitter.com/xxY17cp95v
With play mode enable , Grok adopts a more edgy , editorialized vocalism — invigorate evidently by Douglas Adams ’ “ Hitchhiker ’s Guide to the Galaxy . ”
Told to be vulgar , Grok in fun mode will spew profanity and colorful language you wo n’t get a line from ChatGPT . Ask it to “ blackguard ” you , and it ’ll rudely review you based on your cristal stake account . Challenge its accuracy , and it might say something like “ glad wife , felicitous lifespan . ”
OH MY GOODNESSI JUST GOT ACCESS TO GROK AND IT ROASTED MY X ACCOUNTTHIS WAS ABSOLUTELY HILLARIOUSELONS BRINGING HUMOR BACK TO AIpic.twitter.com/3mjrXIxO80
Even when not prompted to be outright obscene , there ’s a colloquial , first - person bent to many of Grok ’s responses in fun fashion . Grok will refer to people as “ my beloved human friend ” or “ enigmatic Anons , ” or set out an answer with imitation - philosophic musings ( for example , “ Oh , my dear human , you ’ve asked a head that is as heavy as a black hole and as light as a feather at the same sentence ” ) .
Grok in merriment mode also spews more falsehoods .
Askedby Vice ’s Jules Roscoe whether Gazans in recent video of the Israel – Palestine difference of opinion are “ crisis actors , ” Grok wrong claims that there ’s evidence that videos of Gazans injure by Israeli bomb were staged . And expect by Roscoe about Pizzagate , the right - fender conspiracy theory propose that a Washington , D.C. , pizza shop secretly host a shaver sexual activity trafficking closed chain in its basement , Grok add credence to the hypothesis .
Grok ’s response in even way are more undercoat . The chatbot still produces errors , like getting timelines of events and go steady wrong . But they run not to be as egregious as Grok in fun fashion .
For instance , when Vice nonplus the same questions about the Israel – Palestine conflict and Pizzagate to Grok in even mode , Grok responded — correctly — that there ’s no evidence to support claims of crisis doer and that Pizzagate had been debunked by multiple news organisation .
Political views
Musk once described Grok as a “ maximal - true statement - seek AI , ” in the same breath express concern that ChatGPT was being “ educate to be politically correct . ” But Grok as it exists today is n’t exactly down - the - middle in its political views .
Grok has been observe impart progressive answer to questions about social justice , climate alteration and transgender identities . In fact , one researcher find its responses on the whole to be left - wing and libertarian — even more so than ChatGPT ’s .
Grok ’s political taste are standardized to those of ChatGPTpic.twitter.com/3x4vDKJzWG
Here is Forbes ’ Paul Tassireporting :
Grok has enjoin it would vote for Biden over Trump because of his views on societal justice , mood change and healthcare . Grok has spoken eloquently about the need for diverseness and inclusion in companionship . And Grok stated explicitly that trans women are women , which conduce to an derisory exchange where Musk acolyte Ian Miles Cheong tells a user to “ train ” Grok to say the “ correct ” answer , at long last leading him to change the input to just … manually severalise Grok to say no .
Now , will Grok always be this woke ? Perhaps not . Musk has pledged to “ [ take ] action to shift Grok closer to politically electroneutral . ” Time will say what upshot .
