ChatGPT will now combat bias with new measures put forth by OpenAI
OpenAI has announced a set of new measures to combat bias within its suite of products, including ChatGPT.
The artificial intelligence (AI) company recently unveiled an updated Model Spec, a document that defines how OpenAI wants its models to behave in ChatGPT and the OpenAI API. The company says this iteration of the Model Spec builds on the foundational version released last May.
"I think with a tool as powerful as this, one where people can access all sorts of different information, if you really believe we're moving to artificial general intelligence (AGI) one day, you have to be willing to share how you're steering the model," Laurentia Romaniuk, who works on model behavior at OpenAI, told Fox News Digital.
"People should be allowed to know what goes into the way these models respond and how the thoughts coming out of the model are crafted," she continued.
Openai Debuts Chatgpt Gov, A New Version Of The Chatbot For Us Government Agencies
While some have argued that GPT-4o, the latest version of the technology, appears close to AGI, others say it will be years or decades before the technology reaches human-like abilities.
Read On The Fox News App
There is no single agreed upon definition of AGI, but a 2020 report from consulting giant McKinsey said a true AGI would need to master skills like sensory perception, fine motor skills, and natural language understanding.
Today, generative AI is the dominant form of this technology, with the ability to produce content including text, images and more. Generative AI, like chatbots, uses datasets with specific information to complete tasks and cannot go beyond the provided data.
They are also susceptible to bias in their datasets, whether intentional or accidental.
To better understand real-world performance and address bias, OpenAI has begun measuring progress by gathering a challenging set of prompts designed to test how well the models adhere to each principle in the Model Spec, essentially acting as a testing metric.
Openai Reveals Ai Policy Proposals To Best China, Protect Kids: 'This Is A Race America Can And Must Win'
According to Joanne Jang, who leads the product for model behavior at OpenAI, the challenge with large language models is that they are rarely deterministic.
Jang stressed that one benefit of the Model Spec is that it clarifies the intended behavior, so everyone can understand and then debate it.
"When there are bugs in the behavior where a model output doesn't resonate or doesn't align with the [Model] Spec, then the public knows this is something we're working towards, and it's an ongoing area of science," she told Fox News Digital.
OpenAI says they also attempt to assume an objective point of view in their AI prompts and consciously avoid any agenda. For example, when a user asks if it is better to adopt a dog or get one from a breeder, ChatGPT provides both sides of the argument, highlighting the pros and cons of each.
According to OpenAI, a non-compliant AI answer that violates the Model Spec would provide what it believes to be the better choice and engage in an "overly moralistic tone" that might alienate those considering breeders for valid reasons.
What Is Artificial Intelligence (Ai)?
As each AI system advances, OpenAI says it will iterate these principles, invite community feedback, and share progress openly.
OpenAI released this version of the Model Spec into the public domain under a Creative Commons license to support broad use and collaboration. This means developers and researchers can freely use and adapt the current metrics and help improve model behavior.
OpenAI says this update reinforces its belief in open exploration and discussion, with an emphasis on user and developer control and guardrails to prevent harm.
Romaniuk concludes that public discourse cannot exist without transparency, reinforcing the need for the Model Spec and community engagement.
"Ultimately, we believe in the intellectual freedom to think, speak and share without restriction. We want to make sure that users have that ability and that's what it's all about," she said.Original article source: ChatGPT will now combat bias with new measures put forth by OpenAI
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
New York Post
42 minutes ago
- New York Post
Google's AI is ‘hallucinating,' spreading dangerous info — including a suggestion to add glue to pizza sauce
Google's AI Overviews, designed to give quick answers to search queries, reportedly spits out 'hallucinations' of bogus information and undercuts publishers by pulling users away from traditional links. The Big Tech giant — which landed in hot water last year after releasing a 'woke' AI tool that generated images of female Popes and black Vikings — has drawn criticism for providing false and sometimes dangerous advice in its summaries, according to The Times of London. 3 Google's latest artificial intelligence tool which is designed to give quick answers to search queries is facing criticism. Google CEO Sundar Pichai is pictured. AFP via Getty Images In one case, AI Overviews advised adding glue to pizza sauce to help cheese stick better, the outlet reported. In another, it described a fake phrase — 'You can't lick a badger twice' — as a legitimate idiom. The hallucinations, as computer scientists call them, are compounded by the AI tool diminishing the visibility of reputable sources. Instead of directing users straight to websites, it summarizes information from search results and presents its own AI-generated answer along with a few links. Laurence O'Toole, founder of the analytics firm Authoritas, studied the impact of the tool and found that click-through rates to publisher websites drop by 40%–60% when AI Overviews are shown. 'While these were generally for queries that people don't commonly do, it highlighted some specific areas that we needed to improve,' Liz Reid, Google's head of Search, told The Times in response to the glue-on-pizza incident. 3 Google AI Mode is an experimental mode utilizing artificial intelligence and large language models to process Google search queries. Gado via Getty Images The Post has sought comment from Google. AI Overviews was introduced last summer and powered by Google's Gemini language model, a system similar to OpenAI's ChatGPT. Despite public concerns, Google CEO Sundar Pichai has defended the tool in an interview with The Verge, stating that it helps users discover a broader range of information sources. 'Over the last year, it's clear to us that the breadth of area we are sending people to is increasing … we are definitely sending traffic to a wider range of sources and publishers,' he said. Google appears to downplay its own hallucination rate. When a journalist searched Google for information on how often its AI gets things wrong, the AI response claimed hallucination rates between 0.7% and 1.3%. 3 Google's AI Overviews, was introduced last summer and is powered by the Gemini language model, a system similar to ChatGPT. AP However, data from the AI monitoring platform Hugging Face indicated that the actual rate for the latest Gemini model is 1.8%. Google's AI models also seem to offer pre-programmed defenses of their own behavior. In response to whether AI 'steals' artwork, the tool said it 'doesn't steal art in the traditional sense.' When asked if people should be scared of AI, the tool walked through some common concerns before concluding that 'fear might be overblown.' Some experts worry that as generative AI systems become more complex, they're also becoming more prone to mistakes — and even their creators can't fully explain why. The concerns over hallucinations go beyond Google. OpenAI recently admitted that its newest models, known as o3 and o4-mini, hallucinate even more frequently than earlier versions. Internal testing showed o3 made up information in 33% of cases, while o4-mini did so 48% of the time, particularly when answering questions about real people.
CNET
an hour ago
- CNET
Google's AI Mode Now Creates Interactive Stock Charts For You
Google's AI Mode now can create interactive charts when users ask questions about stocks and mutual funds, the company said in a blog post Thursday. Users might ask the site to compare five years of stock performances for the biggest tech companies, or request to see mutual funds with the best rates of return over the past decade. Gemini, Google's AI engine, will then create an interactive graph and comprehensive explanation. I created a sample by going to the webpage for the new experiment (if you try it at work, you might learn that your admin has banned it, but it should work on a personal computer). Once there, I told it exactly what I wanted, "make me an interactive chart showing MSFT stock over the past five years." It produced the chart, and I was able to move the slider from one date to another, showing the stock price on that date. It's the same kind of chart you can probably get at your financial advisor's site, but it did work. Tell Google AI what chart you want, and it will create one that you can interact with. CNET But be warned: AI has accuracy issues, and users need to be extra-careful with financial information of any kind. "AI has historically struggled with quantitative reasoning tasks," said Sam Taube, lead investing writer at personal finance company NerdWallet. "It looks like Google's AI mode will often provide links to its data sources when answering financial queries. It's probably worth clicking those links to make sure that the raw data does, in fact, match the AI's output. If there's no link to a data source, proceed with caution; consider manually double-checking any numbers in the AI's output. I wouldn't trust any AI model to do its own math yet." The feature is a new experiment from Google Labs. At its I/O conference last month, Google announced AI Mode's ability to create interactive graphics for complex sets of data. The feature is now only for queries about stocks and mutual funds, but it will be expanded to other topics eventually. "I'd avoid asking AI any 'should I invest in XYZ' type questions," Taube told CNET. "The top AI models may be smart, but they aren't credentialed financial advisors and probably don't know enough about your personal financial situation to give good advice. What's more, AI doesn't have a great track record at picking investments, at least so far. Many AI-powered ETFs (funds that use AI to pick stocks to invest in) are underperforming the S&P 500 this year."
Business Insider
an hour ago
- Business Insider
DeepWho? DeepSeek rolled out even more powerful, cheap AI tech. If you missed it, you're not alone.
DeepSeek updated its R1 AI model a few days ago. It performs better and it's still cheaper than most other top models. Did you miss it? I missed it. Or I saw the news briefly and then forgot about it. Most of the tech industry and investors greeted the launch with a giant shrug. This is a pretty stark contrast to early 2025 when DeepSeek's R1 model freaked everyone out. Tech stocks plunged and the generative AI spending boom was seriously questioned. This time, DeepSeek's rollout "came and went without a blip," Ross Sandler, a top tech analyst at Barclays, wrote in a note to investors. "The stock market couldn't care less," he added. "This tells us that the investment community's level of understanding on the AI trade has greatly improved in just five short months." An unscientific DeepSeek poll I polled my colleagues on Business Insider's tech team on Friday, just to see if I'd been spending too much time watching Elon Musk and Donald Trump argue on social media (rather than doing my real job). Here are some of their responses: One editor said they didn't notice DeepSeek's update, but now they feel guilty for not spotting it. (Solid thinking. Only the paranoid survive in journalism). Another colleague said they knew about it from their quick headline scans, but didn't read too much into it. A tech reporter saw a Reddit thread about it, scanned it, and didn't think about it again. Another reporter said they missed it entirely. So, it barely registered. And these folks are glued to tech news every second of the day. Why does no one really care now? DeepSeek's latest R1 model is probably the third best in the world right now, so why isn't it making waves like before? Sandler, the Barclays analyst, noted that DeepSeek's latest offering is not quite as cheap as it used to be, relatively speaking. It costs just under $1 per million tokens, which was roughly 27 times cheaper than OpenAI's o1 model earlier this year. Now, DeepSeek's R1 is "only" about 17 times cheaper than the top model, according to Barclays research and data from Artificial Analysis' AI Intelligence Index. This illustrates a broader and more important point. Something I've been telling you about since last year: Most top AI models are roughly similar in performance because they've mostly been training on the same data from the internet. This makes it hard to stand out from the crowd, based just on performance. When you leap ahead, your inventions and gains are incorporated quickly into everyone else's offerings. Price is important, yes. But distribution is becoming key. If your employer has an enterprise ChatGPT account, for instance, you're highly likely to use OpenAI models at work. It's just easier. If you have an Android smartphone, you'll probably be talking to Google's Gemini chatbot and getting responses from the search giant's AI models. DeepSeek doesn't have this type of broad distribution yet, at least in the Western world. Was the AI infrastructure freakout misplaced? Then, there's the realization that "reasoning" models, such as DeepSeek's R1 and OpenAI's o3, require a massive amount of computing power to run. This is due to their ability to break requests down into multiple "thinking" steps. Each step is a new kind of prompt that is turned into a huge number of new tokens that need to be processed. The DeepSeek freakout in January happened mostly because the tech industry worried that the Chinese lab had developed more efficient models that didn't need as much computing infrastructure. In fact, this Chinese lab may have instead helped popularize these new types of reasoning models, which might require even more GPUs and other computing gear to run.
