It’s an open secret that goes up in smoke. OpenAI has just announced at the World Economic Forum (WEF) in Davos that it is developing cybersecurity capabilities for the US Department of Defense.
“Our partnership with the Pentagon includes the development of open source cybersecurity software,” said Anna Makanjoo, global vice president of OpenAI.
OpenAI previously banned the use of AI models and related technologies for “military and war” applications and the creation of malware in ChatGPT’s internal policies.
Questioning ChatGPT Usage Policies
But the company recently removed mention of “creating malware,” “military and war” applications, and “plagiarism” among other things.
The rise of Generative Artificial Intelligence also concerns China.
Chinese scientists have in fact claimed to be working on the formation of a military system based on Baidu’s Ernie Bot. The Chinese technology giant on its side refutes any link with this research.
The Libyan War dataset
However, the South China Morning Post reported last week that a university linked to China’s People’s Liberation Army (PLA) was training the army’s AI system using Baidu’s Large Language Model (LLM). .
Citing an academic article published last month in the Chinese journal Command Control & Simulation, the SCMP indicates that other LLMs have also been used, including GPT3.5 and GPT4. The LLM was then asked to generate combat strategies and results based on the training data.
In the article, the researchers reference the US military’s invasion of Libya in 2011. They explain that they used combat tactics employed by both sides of the conflict, and that they seized desired targets for Ernie Bot to predict US Army strategy.
Rise of Chinese AI
According to the research team, simulation results could facilitate decision-making and be used to train and improve knowledge and cognitive understanding of LLM. The goal is to improve AI-powered military systems and enable them to better understand human intentions, which is essential because human adversaries can be unpredictable.
The researchers added that their work was in its early stages and was being conducted for research purposes only.
Baidu immediately refuted any connection with this affair. In a statement Baidu said that Ernie Bot is available and can be used by the general public.
Baidu added that it had “not engaged in any commercial collaboration or provided a bespoke service” to the team that wrote the academic report or any institution with which the researchers are affiliated.
Baidu said: “Baidu is committed to operating its AI-related products and businesses in compliance with applicable laws and regulations and corporate best practices.”
China also present in quantum computing
Last August, Baidu made Ernie Bot available to the general public in hopes of collecting user feedback to improve the platform. It also plans to release native AI applications that it says would deliver four key GenAI capabilities:
- Reasoning
- Memory
- Generation
- Understanding.
Ernie Bot can generate text, images, audio and video from a prompt, and can speak in several Chinese dialects. Since its introduction with limited access last March, the AI model has gained proficiency in more than 200 types of writing, increasing content quality by 1.6 times, according to Baidu.
Five plugins were also added to Ernie Bot this month, including Baidu Search, Data Analysis and Visualization, and Text to Video Conversion.
Separately, China’s third-generation quantum computer Origin Wukong has processed 33,871 tasks for users around the world since it became operational on Jan. 6, state-run Xinhua reported.
Users from more than 60 countries have remotely accessed the quantum computer at least 350,000 times, with U.S. users leading in frequency, according to the Anhui Quantum Computing Engineering Research Center. The quantum computer runs on a local 72-qubit chip.
Source: “ZDNet.com”