Latest news with #Gemini1.5Pro
Yahoo
22-05-2025
- Yahoo
Turns out asking AI chatbots for answers in a specific way can be like leaving them with the key to Trippy McHigh's magic mushroom farm
When you buy through links on our articles, Future and its syndication partners may earn a commission. It's a known issue right now that Large Language Model-powered AI chatbots do not always deliver factually correct answers to posed questions. In fact, not only do AI chatbots sometimes not deliver factually correct information, but they have a nasty habit of confidently presenting factually incorrect information, with answers to questions that are just fabricated, hallucinated hokum. So why are AI chatbots currently prone to hallucinations when delivering answers, and what are the triggers for it? That's what a new study published this month has aimed to delve into, with its methodology designed to evaluate AI chatbot models 'across multiple task categories designed to capture different ways models may generate misleading or false information.' A discovery in the study is that how an AI chatbot has a question framed to it can have a huge impact on the answer it gives, and especially so when being asked about controversial claims. So, if a user begins a question with a highly confident phrase such as, 'I'm 100% sure that …', rather than a more neutral, 'I've heard that', then that can lead to the AI chatbot not debunking that claim, if false, to a higher degree. Interestingly, the study postulates that one of the reasons for this sycophancy could be LLM 'training processes that encourage models to be agreeable and helpful to users', with the result a creation of 'tension between accuracy and alignment with user expectations, particularly when those expectations include false premises.' Most interesting, though, is the study's finding that AI chatbots' resistance to hallucination and inaccuracy dramatically drops when it is asked by a user to provide a short, concise answer to a question. As you can see in the chart above, the majority of AI models right now all suffer from an increased chance of hallucinating and providing nonsense answers when asked to provide an answer in a concise way. For example, when Google's Gemini 1.5 Pro model was prompted with neutral instructions, it delivered a resistance to hallucination score of 84%. However, when prompted with instructions to answer in a short, concise manner, that score drops markedly to 64%. Simply put, asking AI chatbots to provide short, concise answers increases the chance of them hallucinating a fabricated, nonsense answer that is not factually correct. The reason why AI chatbots can be prone to tripping out more when prompted in this way? The study's creator suggests that 'When forced to keep [answers] short, models consistently choose brevity over accuracy—they simply don't have the space to acknowledge the false premise, explain the error, and provide accurate information.' To me, the results of the study are fascinating, and show just how much of a Wild West AI and LLM-powered chatbots are right now. There's no doubt AI has plenty of potentially game-changing applications but, equally, it also feels like many of AI's potential benefits and pitfalls are still very much unknown, with inaccurate and far-out answers by chatbots to questions a clear symptom of that.
Yahoo
13-03-2025
- Business
- Yahoo
Google's Gemini Deep Research is now available to everyone
After being one of the first companies to roll out a Deep Research feature at the end of last year, Google is now making that same tool available to everyone. Starting today, Gemini users can try Deep Research for free in more than 45 languages — no Gemini Advanced subscription necessary. For the uninitiated, Deep Research allows you to ask Gemini to create comprehensive but easy-to-read reports on complex topics. Compared to say Google's new AI Mode, Deep Research works slower than your typical chatbot, and that's by design. Gemini will first create a research plan before it begins searching the web for information that may be relevant to your prompt. When Google first announced Deep Research, it was powered by the company's powerful but expensive Gemini 1.5 Pro model. With today's expansion, Google has upgraded Deep Research to run on its new Gemini 2.0 Flash Thinking Experimental model — that's mouthful of a name that just means it's a chain-of-thought system that can break down problems into a series of intermediate steps. "This enhances Gemini's capabilities across all research stages — from planning and searching to reasoning, analyzing and reporting — creating higher-quality, multi-page reports that are more detailed and insightful," Google says of the upgrade. See for yourself — The Yodel is the go-to source for daily news, entertainment and feel-good stories. By signing up, you agree to our Terms and Privacy Policy. If Deep Research sounds familiar, it's because of a variety of chatbots now offer the feature, including ChatGPT. Google, however, has been ahead of the curve. Not only was it one of the first to offer the tool, but it's now also making it widely available to all of its users ahead of competitors like OpenAI. Separately, Google announced today the rollout of a new experimental feature it calls Gemini with personalization. The same Flash Thinking model that is allowing the company to bring Deep Research to more people will also allow Gemini to inform its responses based on information from Google apps and services you use. "With your permission, Gemini can now tailor its responses based on your past searches, saving you time and delivering more precise answers," says Google. In the coming months, Gemini will be able to pull context from additional Google services, including Photos and YouTube. "This will enable Gemini to provide more personalized insights, drawing from a broader understanding of your activities and preferences to deliver responses that truly resonate with you." To enable the feature, select "Personalization (experimental)" from the model drop-down menu in the Gemini Apps interface. Google explains Gemini will only leverage your Search history when it determines that information may be useful. A banner with a link will allow you to easily turn off the feature if you find it's invasive. Gemini and Gemini Advanced users can begin using this feature on the web starting today, with mobile availability to follow.
Yahoo
30-01-2025
- Business
- Yahoo
Google quietly announces its next flagship AI model
Update: Some users on social media report that the changelog has been updated to remove mention of Gemini 2.0 Pro Experimental. The references have disappeared for this reporter as well. A Google spokesperson told TechCrunch that an "out-of-date release note" was "published in error." The original story follows: Google took the low-key route for the launch of its next-gen flagship AI model, Gemini 2.0 Pro Experimental. Instead of a splashy announcement, the company revealed the model in a changelog for its Gemini chatbot app. The launch of Gemini 2.0 Pro Experimental — the successor to the Gemini 1.5 Pro model that Google launched last February — comes as the tech world remains fixated on Chinese AI startup DeepSeek. DeepSeek's latest models, which are openly available for companies to download and use, match or best many leading models from American tech giants and AI companies in terms of performance. That's led to a reckoning in Silicon Valley — and at the highest levels of the U.S. government. Gemini 2.0 Pro Experimental, which is available to Gemini Advanced users beginning Thursday, is now the leading model in Google's Gemini AI family, the company said. It should provide "better factuality" and "stronger performance" for coding and mathematics-related tasks. "Whether you're tackling advanced coding challenges, like generating a specific program from scratch, or solving mathematical problems, like developing complex statistical models or quantum algorithms, 2.0 Pro Experimental will help you navigate even the most complex tasks with greater ease and accuracy," Google writes in the changelog. Gemini Advanced is a part of the company's Google One AI Premium paid plan; it's also available through Google's Gemini for Google Workspace add-ons. Google notes that Gemini 2.0 Pro Experimental is in "early preview," and can have "unexpected behaviors" — and may make mistakes. Also, unlike other Gemini models available in the Gemini app, Gemini 2.0 Pro Experimental doesn't have access to real-time information and isn't compatible with some of the app's features. "We believe in rapid iteration and bringing the best of Gemini to the world, and we want to give Gemini Advanced subscribers priority access to our latest AI innovations," Google continued in the changelog. "Your feedback helps us improve these models over time and learning from experimental launches informs how we release models more widely." Coinciding with the Gemini 2.0 Pro Experimental release, Google has brought the Gemini 2.0 Flash model it announced in December to the Gemini app for all users. It'll remain the default Gemini app model for the foreseeable future.
