Topics
Latest
AI
Amazon
Image Credits:Anthropic
Apps
Biotech & Health
clime
Cloud Computing
Commerce
Crypto
endeavor
EVs
Fintech
Fundraising
Gadgets
stake
Government & Policy
Hardware
layoff
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
Security
Social
infinite
Startups
TikTok
deportation
Venture
More from TechCrunch
effect
Startup Battlefield
StrictlyVC
newssheet
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
touch Us
TheOpenAI geartrain wreckmay be playing out in dumb motion before our eyes , but the ship’s company ’s competition is n’t sit around gawking . Anthropic just releasedClaude 2.1,an melioration on its flagship large language role model that keep it competitive with the GPT series — and now has the utilitarian added feature of “ being developed by a companionship not actively at warfare with itself . ”
This new update to Claude has three major improvements : linguistic context window , accuracy and extensibility .
On the context window front , meaning how much data point the mannikin can pay aid to at once , Anthropic has leapfrogged OpenAI : The embattle Sam Altman declare a 128,000 - token window back at the troupe ’s Dev Day ( seems so long ago ! ) , and Claude 2.1 now can handle 200,000 tokens . That ’s enough for “ integral codebases , fiscal statements like S-1s , or even long literary work like The Iliad , ” the company wrote .
Of naturally , having more info does n’t necessarily mean the fashion model handles it as well . GPT-4 is still the gold standard on code generation , for instance , and Claude will handle requests differently than its competitors , some better , some worse . It ’s all a workplace in progress , and ultimately up to users to figure out how best to manage this new capacity .
How the OpenAI debacle could bolster Meta and the ‘ open AI ’ movement
Accuracy also supposedly gets a encouragement ( this is a notoriously difficult concept to quantify ) , grant to “ a big band of complex , factual questions that poke into know weakness in current role model . ” The results show that Claude 2.1 pee fewer incorrect answers , is less probable to hallucinate , and is better at estimate when it ca n’t be certain — the fashion model is “ significantly more likely to except rather than provide incorrect information . ” Again , how useful this is in pattern can only be evaluated by users putting it to sour .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
Lastly , Claude 2.1 can now utilise tools , just like crows and bonobos . No sharp sticks for the LLM , however : It ’s more like the agent functionality we ’re get word issue in models meant to interact with web interfaces . If the model finds that its best move for a question is n’t to reason it out but to simply use a calculator , or a know API , it will do that instead .
For instance , if it does n’t make out which machine or laptop to recommend for someone postulate for Cartesian product advice , it can call out to a model or database better equipped to answer that question , or even do a web lookup if that ’s appropriate .
These iterative improvements will for certain be welcomed by the developer who utilize Claude regularly , and show that every day at OpenAI that ’s lost to might struggles is potentially one miss to the rivalry . Anthropic ’s good example may not always stand toe - to - toe with OpenAI ’s , but this manufacture movesfast . A few free weeks to catch up might make more difference of opinion than anyone expects .
