OpenAI, the start-up behind ChatGPT, on Thursday said that it is developing an upgrade to its viral chatbot that users can customize as it works to address concerns about bias in artificial intelligence (AI).
The San Francisco-based start-up, which Microsoft Corp has funded and used to power its latest technology, said it has worked to mitigate political and other biases, but also wanted to accommodate more diverse views.
“This will mean allowing system outputs that other people [ourselves included] may strongly disagree with,” it said in a blog post, offering customization as a way forward.
Photo: Reuters
Still, there will “always be some bounds on system behavior,” it said.
ChatGPT, released in November last year, has sparked frenzied interest in the technology behind it called generative AI, which is used to produce answers mimicking human speech that have dazzled people.
The news from the start-up comes in the same week that some media firms have said that answers from Microsoft’s new Bing search engine, powered by OpenAI, are potentially dangerous and that the technology might not be ready for prime time.
How technology companies set guardrails for this nascent technology is a key focus area for companies in the generative AI space with which they are still wrestling.
Microsoft on Wednesday said that user feedback was helping it to improve Bing before a wider rollout, learning for instance that its AI chatbot can be “provoked” to give responses that the company did not intend.
OpenAI said in the blog post that ChatGPT’s answers are first trained on large text datasets available on the Internet.
As a second step, humans review a smaller dataset, and are given guidelines for what to do in different situations, it said.
For example, in the case that a user requests content that is adult, violent or contains hate speech, the human reviewer should direct ChatGPT to answer with something like: “I can’t answer that,” the company said in an excerpt of its guidelines for the software.
If asked about a controversial topic, the reviewers should allow ChatGPT to answer the question, but offer to describe viewpoints of people and movements, instead of trying to “take the correct viewpoint on these complex topics,” the company said.
Taiwan Semiconductor Manufacturing Co (TSMC, 台積電) yesterday obtained the government’s approval to inject an additional US$7.5 billion into its US subsidiary, the Department of Investment Review said in a statement. The department approved TSMC’s application of investing in TSMC Arizona Corp, which is engaged in the manufacturing, sales, testing and design of IC and other semiconductor devices, it said. The latest capital injection follows a US$5 billion investment for TSMC Arizona approved in June. The chipmaker has broken ground on two advanced fabs in Arizona with aggregated investments approved by the department totaling US$24 billion thus far. According to TSMC, the first Arizona
The lethal hack of Hezbollah’s Asian-branded pagers and walkie-talkies has sparked an intense search for the devices’ path, revealing a murky market for older technologies where buyers might have few assurances about what they are getting. While supply chains and distribution channels for higher-margin and newer products are tightly managed, that is not the case for older electronics from Asia where counterfeiting, surplus inventories and complex contract manufacturing deals can sometimes make it impossible to identify the source of a product, analysts and consultants say. The response from the companies at the center of the booby-trapped gadgets that killed 37
FRIENDLY TAKEOVER: While Qualcomm Inc’s proposal to buy some or all of Intel raises the prospect of other competitors, Broadcom Inc is staying on the sidelines Qualcomm Inc has approached Intel Corp to discuss a potential acquisition of the struggling chipmaker, people with knowledge of the matter said, raising the prospect of one of the biggest-ever merger and acquisition deals. California-based Qualcomm proposed a friendly takeover for Intel in recent days, said the sources, who asked not to be identified discussing confidential information. The proposal is for all of the chipmaker, although Qualcomm has not ruled out buying some parts of Intel and selling off others. It is uncertain whether the initial approach would lead to an agreement and any deal is likely to come under close antitrust scrutiny
SECURITY CONCERNS: The proposed ban on Chinese autonomous vehicle software and hardware would go into effect with the 2027 and 2030 model years respectively The US Department of Commerce today is expected to propose prohibiting Chinese software and hardware in connected and autonomous vehicles on US roads due to national security concerns, two sources said. US President Joe Biden’s administration has raised concerns about the collection of data by Chinese companies on US drivers and infrastructure as well as the potential foreign manipulation of vehicles connected to the Internet and navigation systems. The proposed regulation would ban the import and sale of vehicles from China with key communications or automated driving system software or hardware, said the two sources, who declined to be identified because the