OpenAI's Latest Response to Claims of Degrading Performance

Today:

OpenAI product leader denies claims GPT-4 has gotten ‘lazier and dumber’

OpenAI, the tech giant known for its artificial intelligence creations, is facing a bunch of legal headaches, including a Federal Trade Commission investigation and a copyright lawsuit from comedian Sarah Silverman. To top it all off, folks on their developer forum have been buzzing about perceived quality drops in OpenAI's GPT-4 model, which powers ChatGPT. Some users reckon the AI's gone from star pupil to back-of-the-class, with theories ranging from tweaked learning algorithms to changed training data.

But hold up, says OpenAI's VP of Product, Peter Welinder. In a tweet, he claims that GPT-4 hasn't gotten "lazier and dumber". In fact, each new model's brainier than the last. He believes that the more users work the system, the more they spot things they didn't catch before. And if folks think the system's slipped, he wants examples to check out.

Still, some folks disagree, with one person saying GPT-4's straight-up worse. And the real stick in the mud? GPT-4's a mystery machine, with developers in the dark about any changes to the model. So for now, the debate about GPT-4's smarts rages on.

READ THE ARTICLE ON VENTUREBEAT

Indian tech giant Wipro will invest $1 billion in AI, including training all staff

Indian software powerhouse, Wipro, is putting down a cool billion on AI over the next three years. They're aiming to make sure every one of their 250,000 workers worldwide knows their way around AI. For the next year, they're hosting crash courses on AI basics and how to use it responsibly, with more tailored training down the line for folks in AI-focused jobs.

Wipro is a major player in India's IT and consulting field. The emergence of generative AI - that's the tech that powers things like this ChatGPT you're using - is shaking up the world, and they're keeping up with the trend.

Wipro's CEO, Thierry Delaporte, sees a big change coming thanks to generative AI. As part of their AI push, Wipro's rolling out software to incorporate AI into all their platforms and tools, whether they're for internal use or client offerings.

AI is quickly becoming a go-to for businesses looking to beef up or replace human-led tasks. Just this week, an Indian startup CEO axed about 90% of his support staff, saying a new AI chatbot could handle customer service quicker than human employees.

READ THE ARTICLE ON CNN

Google leaves Canada out of AI chatbot launch

Canada's been left out in the cold by Google who decided not to launch its AI chatbot, Bard, there. The bot's now available in almost every other country. The company's beef with Canada's Liberal government over new rules is likely the cause.

Bard, Google's own version of OpenAI's ChatGPT, helps users with stuff like writing letters or planning trips. Its launch was put on hold in the EU over privacy issues, but now it's even available there.

Google and Canada's government have been locking horns over the Online News Act. This new law would make Google share dough with Canadian news outlets whose content they use. To sidestep this, Google plans to block news from their platforms in Canada.

READ THE ARTICLE ON NATIONAL POST

Real-time deepfake detection: How Intel Labs uses AI to fight misinformation

Deepfakes, or artificial videos and images, can be used for harmful stuff like spreading false news or tricking people. To tackle this, Intel Labs has created a tool to spot deepfakes as they happen. They use a fancy process that can spot tiny changes in a person's skin color that come with every heartbeat – something a deepfake can't replicate.

This tech is now being used in many places, from newsrooms to social media sites, to stop deepfakes from spreading. Despite the potential for misuse, deepfakes can also be used for good. For instance, people can use deepfake technology to hide their real face online for privacy.

Intel is also making sure this technology is used the right way. They've got a team that includes lawyers, social scientists, and psychologists who look for any potential issues, like bias or harm. With deepfakes getting more common and real-looking, technologies like Intel's could play a big role in keeping us safe.

READ THE ARTICLE ON ZDNET

Nevia Bio is using AI to early detect “silent killer” ovarian cancer

Nevia Bio, co-founded by Dr. Inbal Zafir-Lavie and Dr. Shlomit Yehudai-Reshef, is applying AI for early ovarian cancer detection. The idea? To analyze vaginal secretions, which Zafir-Lavie claims is as close as they can get to the ovaries without surgery.

Most women get diagnosed with ovarian cancer in late stages due to lack of symptoms, resulting in low survival rates - earning the disease the nickname "silent killer". Current detection methods are late-blooming blood tests.

Nevia Bio's tech aims to change this. The plan is to get healthcare providers to offer their tests routinely, but eventually, they hope to offer a home-use kit, making it as easy as ordering off Amazon. This could open doors for women who can't afford frequent doctor visits or live too far from one.

Right now, they're kick-starting a U.S. clinical trial and chasing after FDA approval. They're also looking into using their tech to spot other diseases, like endometrial cancer.

READ THE ARTICLE ON CTECH

This CEO replaced 90% of support staff with an AI chatbot

The CEO of Indian startup Dukaan, Summit Shah, axed 90% of his support staff after introducing an AI-powered chatbot that can field customer questions faster. He said on Twitter that the chatbot, whipped up in just two days, could reply instantly, beating his team's average response time of almost two minutes. The chatbot also cut down problem-solving time by nearly 98%.

This job cull was a "tough but necessary" move, according to Shah. He admitted the layoffs were made last September, around the time when Dukaan shifted its focus from small businesses to consumer-facing brands, reducing the need for live chats or calls.

Shah mentioned that the new tech sliced the cost of customer support by about 85%. It also sorted out a bunch of issues they'd been having, like slow responses and lack of staff when they needed them most.

Despite these layoffs, Dukaan is still on the lookout for fresh talent for several roles in engineering, marketing, and sales. Shah imagines a future where AI and humans play to their strengths, and he's keen to bring AI into graphic design, illustration, and data science.

The announcement of these layoffs has amped up fears about AI stealing jobs, especially after the public release of AI chatbot ChatGPT by Open AI eight months ago. Its ability to provide detailed answers has folks worried it might boot people out of their jobs. The Organisation for Economic Cooperation and Development warns that not just low- and middle-skilled jobs, but also professions like finance, medicine, and law, are at risk of being automated by AI.

READ THE ARTICLE ON CNN