Category: News

“Clinicians may also become de-skilled as over-reliance on the outputs of AI diminishes critical thinking,” Shegewi said. “Large-scale deployments will likely raise issues concerning patient data privacy and regulatory compliance. The risk for bias, inherent in any AI model, is also huge and might harm underrepresented populations.”

Additionally, AI’s increasing use by healthcare insurance companies doesn’t typically translate into what’s best for a patient. Doctors who face an onslaught of AI-generated patient care denials from insurance companies are fighting back — and they’re using the same technology to automate their appeals.

“One reason the AI outperformed humans is that it’s very good at thinking about why it might be wrong,” Rodman said. “So, it’s good at what doesn’t fit with the hypothesis, which is a skill humans aren’t very good at. We’re not good at disagreeing with ourselves. We have cognitive biases.”

Of course, AI has its own biases, Rodman noted. The higher ratio of sex and racial biases has been well documented with LLMs, but it’s probably less prone to biases than people are, he said.

Even so, bias in classical AI has been a longstanding problem, and genAI has the potential to exacerbate the problem, according to Gartner’s Walk. “I think one of the biggest risks is that the technology is outpacing the industry’s ability to train and prepare clinicians to detect, respond to, and report these biases,” she said. 

GenAI models are inherently prone to bias due to their training on datasets that may disproportionately represent certain populations or scenarios. For example, models trained primarily on data from dominant demographic groups might perform poorly for underrepresented groups, said Mutaz Shegewi, a senior research director with IDC’s Worldwide Healthcare Provider Digital Strategies group.

“Prompt design can further amplify bias, as poorly crafted prompts may reinforce disparities,” he said. “Additionally, genAI’s focus on common patterns risks overlooking rare but important cases.”

For example, research literature that’s ingested by LLMs is often skewed toward white males, creating critical data gaps regarding other populations, Mutaz said. “Due to this, AI models might not recognize atypical disease presentations in different groups. Symptoms for certain diseases, for example, can have stark differences between groups, and a failure to acknowledge such differences could lead to delayed or misguided treatment,” he said.

With current regulatory structures, LLMs and their genAI interfaces can’t accept liability and responsibility the way a human clinician can. So, for “official purposes,” it’s likely a human will still be needed in the loop for liability, judgement, nuance, and the many other layers of evaluation and support patients need.

Chen said it wouldn’t surprise him if physicians were already using LLMs for low-stakes purposes, like explaining medical charts or generating treatment options  for less-severe symptoms.

“Good or bad, ready or not, Pandora’s box has already been opened, and we need to figure out how to effectively use these tools and counsel patients and clinicians on appropriately safe and reliable ways to do so,” Chen said.

All this, due in large part to the explosion of ChatGPT. In fact, six months after the chatbot’s release, the Future of Life Institute asked for a pause in its development in an open letter, saying its risks could not be controlled, even going so far as to say that it could pose a danger to our civilization as we know it if systems were built that surpassed humans. More than 31,000 people signed the letter, including industry figures such as Apple cofounder Steve Wozniak and OpenAI cofounder Elon Musk.

ChatGPT broke all predictions. A study by UBS found that it was the fastest consumer application to reach 100 million users, in just two months, although it has since been surpassed by Meta’s social network Threads. And, at the business level, it has one million licenses. In total, it has more than 180.5 million monthly active users as of April of this year, and its page was accessed by 1,625 million visitors in the month of February, according to PrimeWeb.

“It has transformed the way we interact with technology,” says Fernando Maldonado, an independent analyst. “Today, anyone can access AI without the need for advanced knowledge or intermediaries, something that was previously reserved for specialists.” 

Focus on market dominance and security practices

The investigation centers on Microsoft’s bundling of office productivity and security software with its cloud services, a practice critics argue disadvantages competitors in the authentication and cybersecurity markets. The FTC is particularly examining Microsoft Entra ID, its user authentication service, amid complaints that licensing terms and integration with its cloud offerings impede rival companies.

Microsoft’s role as a major government contractor and recent cybersecurity incidents involving its products have added urgency to the probe. The company provides billions of dollars in services to US agencies, including the Department of Defense, making its practices critical to national security.

In November 2023, FTC flagged concerns about the concentrated nature of the cloud market, warning that outages or performance issues could ripple through the economy. The authority had then collected feedback from civil society, industry stakeholders, and academia to prepare its report. As per the feedback, the majority of the concerns were related to competition and licensing practices.

In the interests of objectivity, I should say up from I love the new Mac mini. It’s a triumph, a culmination of everything the first Mac mini aimed to be, but much, much better. Introduced along with the also superb MacBook Pro, Apple’s Mac line-up proves that, with Apple Silicon inside, the company is a the top of its game.

What you can expect under the hood

All this capability comes because of the amazing M-series processor Apple has slotted inside and reflects the device’s extensive processor history that straddles the company’s PowerPC chips on its first release, the Intel years, and today’s super-efficient, low-power chips that put Apple ahead of the industry. There’s a lot to love, starting at $599 (though the M4 Pro with 14‑core CPU and 20‑core GPU, 48GB, and 1TB SSD model I tested costs a lot more, $2,199.) That price tag might dent the superlatives a little, but probably not fatally. 

For a company made famous by the quality of its design, the Mac mini you see today isn’t a major departure from the models of yesteryear, other than size. This third major redesign remains faithful to the breed — a compact all-in-one metal box designed to work with the mouse, keyboard and display you already own. Now just 2-in. high, the 5-in.-by-5-in. (100% carbon neutral aluminum) box remains, resolutely, a Mac mini.

Notably, last month Anthropic introduced “Computer Use” to Claude 3.5 Sonnet. This capability allows the model to use computers as people do; it can quickly move cursors, toggle between tabs, navigate websites, click buttons, type, and compile research documents in addition to its generative capabilities. All told, the company claims that Sonnet outperforms all other available models on agentic coding tasks.

Claude has experienced rapid adoption since its addition to Amazon Bedrock, AWS’ fully-managed service for building generative AI models, in April 2023, and now supports “tens of thousands” of companies across numerous industries, according to AWS. The foundation models are used to build a number of functions, including chatbots, coding assistants, and complex business processes.

“This has been a year of breakout growth for Claude, and our collaboration with Amazon has been instrumental in bringing Claude’s capabilities to millions of end users on Amazon Bedrock,” Dario Amodei, co-founder and CEO of Anthropic, said in an announcement.

In connection with this year’s Ignite conference, Microsoft has unveiled a new interpretation tool that will be added to Teams in the spring. What makes the voice cloning tool — currently called “Interpreter In Teams” — special is that users will be able to use your own voice to speak in other languages ​​in real time.

According to Techcrunch, users need a subscription to Microsoft 365 to have access to the technology.

Initially, the tool will support nine languages: English, French, Italian, Portuguese, Spanish, German, Japanese, Korean and Mandarin. More languages ​​are likely to be added over time.

For its part, Perplexity said in an updated FAQ that its web crawler, PerplexityBot, will not index the full or partial text content of any site that disallows it using robots.txt code. Robots.txt files are common simple text files stored on a web server to instruct web crawlers about which pages or sections of a website they are allowed to crawl and index.

“PerplexityBot only crawls content in compliance with robots.txt,” the FAQ explained. Perplexity also said it does not build “foundation models,” (also known as large language models), “so your content will not be used for AI model pre-training.”

The bottom line, Yamin said, is that search engines are in a “tricky position” as genAI evolves. “They want to provide the best results to users, which increasingly involves AI-generated or AI-enhanced content. At the same time, they need to protect original creators and maintain the integrity of search results. We’re seeing efforts to strike this balance, but it’s a complex issue that will take time to fully address.”

OpenAI pretty much concedes this in the report: “In this work, we will sidestep the open-endedness of language models by considering only short, fact-seeking questions with a single answer. This reduction of scope is important because it makes measuring factuality much more tractable, albeit at the cost of leaving open research questions such as whether improved behavior on short-form factuality generalizes to long-form factuality.”

Later in the report, OpenAI elaborates: “A main limitation with SimpleQA is that while it is accurate, it only measures factuality under the constrained setting of short, fact-seeking queries with a single, verifiable answer. Whether the ability to provide factual short answers correlates with the ability to write lengthy responses filled with numerous facts remains an open research question.”

Here are the specifics: SimpleQA consists of 4,326 “short, fact-seeking questions.” 

French tech company Mistral AI has launched a new online moderation tool based on the AI ​​model Ministral 8B that can detect and remove offensive or illegal posts automatically. (There is still a risk of some misjudgments, however.)

According to Techcrunch, for example, some studies have shown that posts about people with disabilities can be flagged as “negative” or “toxic” even though that’s not the case.

Initially, Mistral’s new moderation tool will support Arabic, English, French, Italian, Japanese, Chinese, Korean, Portuguese, Russian, Spanish and German, with more languages ​​are on the way later. Mistral in July launched a large language model that can generate longer tranches of code faster than other open-source models.