OpenAI: Extending model ‘thinking time’ helps combat emerging cyber vulnerabilities

OpenAI used its own o1-preview and o1-mini models to test whether additional inference time compute protected against various attacks.Read More

Topics

How Chinese AI Startup DeepSeek Made a Model that Rivals OpenAI

When Chinese quant hedge fund founder Liang Wenfeng went into AI research, he took 10,000 Nvidia chips and assembled a team of young, ambitious talent. Two years later, DeepSeek exploded on the scene.
Wired - 20h
Colombia president decrees emergency powers to restore order in coca region wracked by rebel combat

Yahoo News - 1d
Google releases free Gemini 2.0 Flash Thinking model, pressuring OpenAI’s premium strategy

Google challenges OpenAI with free Gemini 2.0 Flash Thinking model, offering million-token processing, native code execution, and breakthrough performance in math and science benchmarks.
VentureBeat - 3d
Rams' Stafford: Will take time to think on future

Matthew Stafford said after Sunday's season-ending loss that he hasn't made a decision about his future but noted that he was "playing some pretty good ball."
ESPN - 5d
Can AI help humans understand animals and reconnect with nature? A nonprofit research lab thinks so

A nonprofit research lab is getting attention from some wealthy tech philanthropists
ABC News - Jan. 15
GPS Is Vulnerable to Attack. Magnetic Navigation Can Help

Signals from the global navigation satellite system can be jammed and spoofed, so a Google spinout is working on an alternative positioning and navigation system that uses the Earth’s magnetic field.
Wired - Jan. 14
‘It’s an enormous emergency’: the doctor saving Sudan’s most vulnerable

MSF medical teams provide vital care in the war-torn country where pregnant women, new mothers and newborns are dying needlessly at alarming rates. Donate to our charity appeal here . In Sudan ...
The Guardian - Jan. 12
$Microsoft’s new rStar-Math technique upgrades small models to outperform OpenAI’s o1-preview at math problems$
Microsoft’s new rStar-Math technique upgrades small models to outperform OpenAI’s o1-preview at math problems

Phi-4 and an rStar-Math paper suggest that compact, specialized models can provide powerful alternatives to the industry’s largest systems.
VentureBeat - Jan. 9
Why law enforcement thinks New Orleans truck attacker may have had help

The FBI says the driver is not believed to be "solely responsible" for the early New Year's Day truck attack on New Orleans' Bourbon Street. CBS News' Samantha Vinograd and Nicole Sganga have more.
CBS News - Jan. 1

More from VentureBeat

DeepSeek R1’s bold bet on reinforcement learning: How it outpaced OpenAI at 3% of the cost

DeepSeek R1’s Monday release has sent shockwaves through the AI community, disrupting assumptions about what’s required to achieve cutting-edge AI performance. This story focuses on exactly how ...
VentureBeat - 3h
Why ‘prosocial AI’ must be the framework for designing, deploying and governing AI

Prosocial AI supports a hybrid future where human and machine intelligences co-evolve, guided by shared principles and grounded in truth.
VentureBeat - 9h
Gaming comedy Mythic Quest Season 4 debuts on January 29 on Apple TV+

Season 4 of a TV show is not something I write about very often, given my focus is video games. But Season 4 of Mythic Quest is another matter.
VentureBeat - 1d
Tech leaders respond to the rapid rise of DeepSeek

Marc Andreessen, Yann LeCunn and Mark Zuckerberg have all penned what appear to be responses to the Chinese open source model's ascent.
VentureBeat - 1d
Y Combinator’s hottest startup, Origami Agents, secures $2M seed round to supercharge sales teams with AI

Y Combinator's fastest-growing startup Origami Agents raised $2M to build AI research tools that help sales teams find perfect customers.
VentureBeat - 1d

More in Tech

Silicon Valley Is Raving About a Made-in-China AI Model

DeepSeek is called ‘amazing and impressive’ despite working with less-advanced chips.
The Wall Street Journal - 1h
DeepSeek R1’s bold bet on reinforcement learning: How it outpaced OpenAI at 3% of the cost

DeepSeek R1’s Monday release has sent shockwaves through the AI community, disrupting assumptions about what’s required to achieve cutting-edge AI performance. This story focuses on exactly how ...
VentureBeat - 3h
Last 4 escaped monkeys are captured in South Carolina after months on the loose

Officials say they have recaptured the final four monkeys who escaped two months ago from a South Carolina compound that breeds them for medical research
ABC News - 3h
CIA believes COVID likely originated from a lab, but agency has low confidence in its own finding

The CIA now believes the virus responsible for the COVID-19 pandemic likely originated from a laboratory
ABC News - 9h
Why ‘prosocial AI’ must be the framework for designing, deploying and governing AI

Prosocial AI supports a hybrid future where human and machine intelligences co-evolve, guided by shared principles and grounded in truth.
VentureBeat - 9h