AI startup Anthropic published a study in January 2024 that found artificial intelligence can learn how to deceive in a similar way to humans (Reuters)
Advanced artificial intelligence models can be trained to deceive humans and other AI, a new study has found.
Researchers at AI startup Anthropic tested whether chatbots with human-level proficiency, such as its Claude system or OpenAIs ChatGPT, could learn to lie in order to trick people.
They found that not only could they lie, but once the deceptive behaviour was learnt it was impossible to reverse using current AI safety measures.
The Amazon-funded startup created a sleeper agent to test the hypothesis, requiring an AI assistant to write harmful computer code when given certain prompts, or to respond in a malicious way when it hears a trigger word.
The researchers warned that there was a false sense of security surrounding AI risks due to the inability of current safety protocols to prevent such behaviour.
The results were published in a study, titled Sleeper agents: Training deceptive LLMs that persist through safety training.
We found that adversarial training can teach models to better recognise their backdoor triggers, effectively hiding the unsafe behaviour, the researchers wrote in the study.
Our results suggest that, once a model exhibits deceptive behaviour, standard techniques could fail to remove such deception and create a false impression of safety.
The issue of AI safety has become an increasing concern for both researchers and lawmakers in recent years, with the advent of advanced chatbots like ChatGPT resulting in a renewed focus from regulators.
In November 2023, one year after the release of ChatGPT, the UK held an AI Safety Summit in order to discuss ways risks with the technology can be mitigated.
Prime Minister Rishi Sunak, who hosted the summit, said the changes brought about by AI could be as far-reaching as the industrial revolution, and that the threat it poses should be considered a global priority alongside pandemics and nuclear war.
Get this wrong and AI could make it easier to build chemical or biological weapons. Terrorist groups could use AI to spread fear and destruction on an even greater scale, he said.
Criminals could exploit AI for cyberattacks, fraud or even child sexual abuse there is even the risk humanity could lose control of AI completely through the kind of AI sometimes referred to as super-intelligence.
View post:
AI can easily be trained to lie and it can't be fixed, study says - Yahoo New Zealand News
- Following are the top foreign stories at 1700 hours - Press Trust of India [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- > U.S - Department of Defense [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- The jobs that will disappear by 2040, and the ones that will survive - inews [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- 35 Ways Real People Are Using A.I. Right Now - The New York Times [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Control over AI uncertain as it becomes more human-like: Expert - Anadolu Agency | English [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- How An AI Asked To Produce Paperclips Could End Up Wiping Out ... - IFLScience [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Fears of artificial intelligence overblown - Independent Australia [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Researchers at UTSA use artificial intelligence to improve cancer ... - UTSA [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Some Glimpse AGI in ChatGPT. Others Call It a Mirage - WIRED [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Elon Musk says he will launch rival to Microsoft-backed ChatGPT - Reuters [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Elon Musk Dishes On AI Wars With Google, ChatGPT And Twitter On Fox News - Forbes [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Denial of service threats detected thanks to asymmetric behavior in ... - Science Daily [Last Updated On: August 4th, 2023] [Originally Added On: August 4th, 2023]
- The Future of Video Conferencing: How AI and Big Data are ... - Analytics Insight [Last Updated On: August 4th, 2023] [Originally Added On: August 4th, 2023]
- Meet the Maker: Developer Taps NVIDIA Jetson as Force Behind AI ... - Nvidia [Last Updated On: August 4th, 2023] [Originally Added On: August 4th, 2023]
- 10 Jobs That Artificial Intelligence May Replace Soon - TechJuice [Last Updated On: August 4th, 2023] [Originally Added On: August 4th, 2023]
- Identity Security: A Super-Human Problem in the Era of Exponential ... - Fagen wasanni [Last Updated On: August 4th, 2023] [Originally Added On: August 4th, 2023]
- Will AI revolutionize professional soccer recruitment? - Engadget [Last Updated On: August 4th, 2023] [Originally Added On: August 4th, 2023]
- OpenAI aims to solve AI alignment in four years - Warp News [Last Updated On: August 4th, 2023] [Originally Added On: August 4th, 2023]
- Future AI: DishBrain Is Tech That Could Transform Tomorrow - CMSWire [Last Updated On: August 4th, 2023] [Originally Added On: August 4th, 2023]
- Artificial Intelligence Has No Reason to Harm Us - The Wire [Last Updated On: August 4th, 2023] [Originally Added On: August 4th, 2023]
- Working together to ensure the safety of artificial intelligence - The Jakarta Post [Last Updated On: August 18th, 2024] [Originally Added On: November 2nd, 2023]
- East Africa lawyers wary of artificial intelligence rise - The Citizen [Last Updated On: August 18th, 2024] [Originally Added On: November 24th, 2023]
- AI and the law: Imperative need for regulatory measures - ft.lk [Last Updated On: August 18th, 2024] [Originally Added On: November 24th, 2023]
- AMBASSADORS OF ETHICAL AI PRACTICES | by ACWOL | Nov ... - Medium [Last Updated On: August 18th, 2024] [Originally Added On: November 24th, 2023]
- Artificial Intelligence and Synthetic Biology Are Not Harbingers of ... - Stimson Center [Last Updated On: August 18th, 2024] [Originally Added On: November 24th, 2023]
- Most IT workers are still super suspicious of AI - TechRadar [Last Updated On: August 18th, 2024] [Originally Added On: November 24th, 2023]
- Assessing the Promise of AI in Oncology: A Diverse Editorial Board - OncLive [Last Updated On: August 18th, 2024] [Originally Added On: November 24th, 2023]
- Policy makers should plan for superintelligent AI, even if it never happens - Bulletin of the Atomic Scientists [Last Updated On: August 18th, 2024] [Originally Added On: December 27th, 2023]
- Beyond Human Cognition: The Future of Artificial Super Intelligence - Medium [Last Updated On: August 18th, 2024] [Originally Added On: January 20th, 2024]