[ad_1]
Anthropic used Pokémon to benchmark its newest AI model. Yes, really.
Build a website that pays you back with real revenue.
In a blog post published Monday, Anthropic said that it tested its latest model, Claude 3.7 Sonnet, on the Game Boy classic Pokémon Red. The company equipped the model with basic memory, screen pixel input, and function calls to press buttons and navigate around the screen, allowing it to play Pokémon continuously.
A unique feature of Claude 3.7 Sonnet is its ability to engage in “extended thinking.” Like OpenAI’s o3-mini and DeepSeek’s R1, Claude 3.7 Sonnet can “reason” through challenging problems by applying more computing — and taking more time.
That came in handy in Pokémon Red, apparently.
Compared to a previous version of Claude, Claude 3.0 Sonnet, which failed to leave the house in Pallet Town where the story begins, Claude 3.7 Sonnet successfully battled three Pokémon gym leaders and won their badges.
Now, it’s not clear how much computing was required for Claude 3.7 Sonnet to reach those milestones — and how long each took. Anthropic only said that the model performed 35,000 actions to reach the last gym leader, Surge.
It surely won’t be long before some enterprising developer finds out.
Pokémon Red is more of a toy benchmark than anything. However, there is a long history of games being used for AI benchmarking purposes. In the past few months alone, a number of new apps and platforms have cropped up to test models’ game-playing abilities on titles ranging from Street Fighter to Pictionary.
[ad_2]
Source link
Related posts:
Stay Safe Online: Essential Tips for Safer Internet Day
Is Your Phone Your Best Friend or a Silent Spy?
Wipe Your Digital Footprints with Data Wipe Software
No, you’re not fired – but beware of job termination scams
DeceptiveDevelopment targets freelance developers
Fake job offers target coders with infostealers
Belarus-Linked Ghostwriter Uses Macropack-Obfuscated Excel Macros to Deploy Malware
LightSpy Expands to 100+ Commands, Increasing Control Over Windows, macOS, Linux, and Mobile
CISA Adds Microsoft and Zimbra Flaws to KEV Catalog Amid Active Exploitation
Malicious PyPI Package "automslc" Enables 104K+ Unauthorized Deezer Music Downloads
CERT-UA Warns of UAC-0173 Attacks Deploying DCRat to Compromise Ukrainian Notaries
Three Password Cracking Techniques and How to Defend Against Them
New Linux Malware ‘Auto-Color’ Grants Hackers Full Remote Access to Compromised Systems
SOC 3.0 - The Evolution of the SOC and How AI is Empowering Human Talent
Leaked Black Basta Chat Logs Reveal $107M Ransom Earnings and Internal Power Struggles
Microsoft: Russian-Linked Hackers Using 'Device Code Phishing' to Hijack Accounts
AI-Powered Social Engineering: Ancillary Tools and Techniques
Lazarus Group Deploys Marstech1 JavaScript Implant in Targeted Developer Attacks
New “whoAMI” Attack Exploits AWS AMI Name Confusion for Remote Code Execution
Android's New Feature Blocks Fraudsters from Sideloading Apps During Calls
New Golang-Based Backdoor Uses Telegram Bot API for Evasive C2 Operations
⚡ THN Weekly Recap: Google Secrets Stolen, Windows Hack, New Crypto Scams and More
CISO's Expert Guide To CTEM And Why It Matters
South Korea Suspends DeepSeek AI Downloads Over Privacy Violations
Microsoft Uncovers New XCSSET macOS Malware Variant with Advanced Obfuscation Tactics
Cybercriminals Exploit Onerror Event in Image Tags to Deploy Payment Skimmers
New Xerox Printer Flaws Could Let Attackers Capture Windows Active Directory Credentials
Winnti APT41 Targets Japanese Firms in RevivalStone Cyber Espionage Campaign
Juniper Session Smart Routers Vulnerability Could Let Attackers Bypass Authentication
Debunking the AI Hype: Inside Real Hacker Tactics
Build a website that pays you back with real revenue.
