DeveLearn 的动态

So much to say about OpenAI's GPT-4o release yesterday, but for the moment I'll just point out that OpenAI has forced company's hands in the AI adoption race. By making their advanced models free to everyone, the ability for a company to control access has gone out the window. Previously, free GenAI models were frankly terrible compared to for-fee versions. The majority of your employees' exposure to GenAI was likely to a pretty annoying toy, not an incredibly capable graduate-level assistant. The difference is that now, your entire company has access to the premium version. Every single employee can build their own tools to make themselves look smarter, to do more of their work for them, and to automate their own workflows... all without you knowing. And further, they can share any advancements they make with only those parties they choose to. This leaves corporations in a tough spot. Those that continue to hesitate will see their IP leak more profusely, their top talent exit faster, and their internal processes corrode as employees use these tools in secret. Alternatively, those that adopt AI as broadly as possible face significant security, IP risk, and cultural adoption challenges - but will at least be able to compete. Google's I/O developer conference, happening today, is certain to have some big AI announcements. But at least for now, OpenAI has just shifted the playing field significantly. Again. https://lnkd.in/dcfVrmsk

OpenAI Just Opened the Floodgates for Consumer AI Dan Shipper’s and Evan Armstrong’s key takeaways from Open AI’s new GPT-4o: ?? GPT-4o made a technical leap by combining multiple modalities into one AI system. It can natively process inputs like voice and video without conversion steps. ??? The desktop app is a strategic move to stay ahead of Big Tech. OpenAI can access local data and could eventually operate systems directly.? ?? #OpenAI is doubling down on the "make it free" playbook to maintain its position as the dominant consumer AI platform. ?? Most excitingly, it’s the first step toward generative AI becoming the meta-layer running on top of all our apps and workflows. This consumer play is all about strengthening OpenAI's position before an expected "#GPT-5 moment" of radically smarter #AI. The race to shape our digital experiences is heating up. If you want to stay ahead, follow Every Inc. for the best tech insights on the internet. P.S. Join 70,000+ others who read our newsletter to stay ahead of the tech curve: https://lnkd.in/dPNxW8-Y #AIRace #ConsumerAI

Over the weekend I took a first look at #GoogleGemini, Google's GPT-4 class #LLM (and which replaces 'Bard'). I wanted to take a look at Gemini Advanced ,which is Google's most powerful AI model (powered by Gemini Ultra 1.0) and to gain access required me to sign up to their premium subscription which provides 2 months of free access but which is c. $20 / ￡19 thereafter (similar to #OpenAI's #ChatGPT Plus) and which also features a storage increases as part of the package. Initial thoughts are that its definitely a step up from Bard and a real challenger to the top class AI solutions such as Anthropic's Claude 2.1 and OpenAI's ChatGPT, although Gemini still only has a 32K context window compared to 128K for GPT-4 Turbo and the massive 200K for Claude 2.1. As with Microsoft's copilot, which can interact with the Microsoft ecosystem of applications, Gemini can interact with the Google ecosystem Apps such as Gmail. Docs, Sheets etc (you have to grant it explicit permissions to the your account to be able to do this). Gemini is also multi-modal. You can ask it to describe, or ask questions about images you upload. It is not yet able to produce images (although this is supposed to be coming) and you cannot attach files to use as part of the conversation as you can for example with Claude 2.1. I provided Gemini U. a short piece of Python code that had some encoding issues and asked it to correct it, something I've seen trip up other models (including the prior incarnation of Gemini, Bard), and it did with aplomb, first time, recognizing the issue, and correcting it, immediately. It will be interesting to see how Gemini evolves. It is definitely a top class model but it will likely be defined by how it fits into Google's overall AI vision. Microsoft vision of its AI seems to be centered around copilots whereas OpenAI is moving down the autonomous agent route, similar to what Amazon seem to be doing with Amazon Q. If you want to find out more about Gemini, worth a read is Google's own in-depth analysis of Gemini ? https://lnkd.in/eWEQZC7Q #AI #EnterpriseAI

Is This the End of Google??? Yesterday, OpenAI launched its first search engine prototype, SearchGPT. As they have done with their other products, OpenAI is granting access only to a select few. Their intention is to gather necessary feedback through this prototype, as they plan to integrate it with ChatGPT in the future. For a long time, Google has been the king of internet searches. However, it seems they have fallen behind in innovation, and OpenAI has managed to advance far beyond the tech giant. Here are some key highlights: Highlights: - SearchGPT is specifically a search engine and operates independently of the training of OpenAI's basic generative AI models. Websites can appear in search results even if they choose not to participate in generative AI training. - If you have a website and are interested in appearing on the SearchGPT search engine, check out this link: https://lnkd.in/evzabjZ2. - If you want to be part of the privileged list of users who can test and provide feedback on the new search engine, sign up here: https://chatgpt.com/search. Remember, Perplexity already exists and was created with the same vision as SearchGPT, innovating search engines. Here's the link to try it out: https://www.perplexity.ai/. I hope this information is useful. Have a great weekend! #OpenAI #SearchGPT #Innovation #SearchEngine #AI #TechNews

?? Google's worst nightmare has come true: OpenAI is finally challenging Google Search! ?? ? SearchGPT, powered by GPT-4, can summarize link results with remarkable efficiency. However, there's a catch: it's currently available to only 10,000 users as a prototype. ? The reason? Google's past blunders, like the Google AI overview feature fiasco, have made them cautious. Remember the disaster where AI suggested adding glue to pizza sauce and highlighted the health benefits of tobacco? https://lnkd.in/gpJ6_MXT ? All we can do is join the waitlist and anticipate what's next ?? https://chatgpt.com/search? OpenAI aims to learn from this prototype, refine it, and eventually integrate the technology into ChatGPT. As Sam Altman puts it, the aim is to "make it real-time and maximally helpful." ? The future of search is here: generative AI models combined with real-time web information. Google, take note! ??? #AI #MachineLearning #OpenAI #GPT4 #SearchGPT #Google #TechInnovation #FutureOfSearch #ArtificialIntelligence #TechNews

"If you can understand a spreadsheet, then you can understand #AI," boasts Anand. The 1.25GB spreadsheet has been made available on GitHub for anyone to download and play with. This spreadsheet can handle just 10 tokens of input, which is tiny compared to the 128,000 tokens that GPT-4 Turbo can handle. However, it is still good for a demo and Anand claims that his "low-code introduction" is ideal as an #LLM grounding for the likes of tech execs, marketers, product managers, AI policymakers, ethicists, as well as for developers and scientists who are new to AI.

?? Google in AI reasoning showdown with OpenAI tl;dr: Forget just writing or talking, Google's AI is learning to *think* like us, tackling multi-step problems with human-like logic. They're gunning straight for OpenAI's o1 crown. * * * ?? The Big Picture: The AI battleground just got a whole lot more interesting. AI chatbots aren’t just spitting out statistically correct responses anymore. New AI can actually *reason* - breaking down complex challenges into smaller steps and solving them logically, kind of like we do. Google's been quietly developing this "reasoning" capability in their AI models, using a technique called "chain-of-thought prompting." It essentially similar to what OpenAI has done, and allows the AI to mimic our thought processes, making them significantly more adept at handling intricate tasks in fields like mathematics and computer programming. ?? What it means for You: We’ll soon see AI assistants that can not only answer your questions but actually solve complex problems, strategize, and even innovate alongside you. We're talking about AI that can analyze market trends to identify new opportunities, optimize supply chains with human-like intuition, and even personalize customer experiences on a whole new level. ?? My take on It: This is a crucial step towards developing genuinely useful AI systems. While OpenAI has made impressive progress, Google's focus on reasoning would create competition, and hopefully lead to interesting innovation in this area. It's a clear signal that the future of AI isn't just about bigger models, it's about smarter, more intuitive ones that can truly understand and navigate the complexities of our world. #Innovation #AI #ChainOfThought #TechLeadership #AIReasoning

**Summary: Integration of GPT-4o into Daily Life** **1. Overview of GPT-4o:** - GPT-4o is the latest AI model from OpenAI, launched recently. - It has significant capabilities like real-time translations and environment-based question answering. **2. Sam Altman's Usage:** - **Routine Integration:** - Altman keeps his phone on the table while working. - Uses GPT-4o to get instant responses without interrupting his workflow. - **Benefit:** - Allows multitasking by providing answers and translations quickly. **3. Impact on Personal Life:** - **Privacy Challenges:** - Altman’s fame makes it difficult to maintain privacy in public spaces. - He finds it isolating due to constant recognition. **4. Professional Challenges:** - **Brief Dismissal:** - Experienced a short but intense dismissal from OpenAI. - Described the period as "adrenaline-charged" until he was reinstated. **5. Significance:** - Highlights the personal and professional challenges of leading a major tech company. - Shows how advanced AI models like GPT-4o can enhance productivity in daily routines. **Implementation in Daily Life:** - **Example:** - Place a smart device with GPT-4o on your desk to get quick answers or translations while working. - **Benefit:** - Helps in staying focused on tasks without needing to switch devices or tabs. **Related News:** - Platforms like LinkedIn and TikTok are making changes to handle AI-generated content and fake accounts. - Amazon plans to enhance Alexa to become a more advanced voice assistant, though at a cost.

???????? ???????? ?????????????? ???????????????????? ?????????? ????????????’?? ??????????????? ???????????? ??????-?? , ?????? ?????????????? ?????????? ???? ??????-?? ?????????? ?? ? Is OpenAI about to lose its lead to Google? That’s the question churning in the minds of many people in the industry this week as results from a popular benchmarking website show that Google’s Bard (powered by Gemini Pro) is the first contender to outperform OpenAI’s GPT-4. The numbers come from the Chatbot Arena Leaderboard, a crowdsourced open platform for AI model evaluation that ranks models based on 200,000+ human preference votes. Google’s Bard jumped to the second spot last week with a slender lead over GPT-4 — but still lags behind GPT-4 Turbo. The sudden leap in performance is impressive considering that Bard is currently powered by Gemini Pro — Google’s second-tier AI model which is one step below Gemini Ultra, the top-end model which is coming soon. The step-up in performance has raised speculation that Gemini Ultra could claim the GPT-4 Turbo’s crown as the top-performing model very soon. But AI models are a game of cat and mouse. OpenAI has consistently beaten its own records and others are catching up. Speculation has been rife for weeks that OpenAI is close to releasing GPT-5, which could reclaim the top spot from the supposedly superior Gemini Ultra. Meta is also rumored to be on the verge of announcing Llama 3, the next version of its popular open-source AI model. Some observers have gone as far as to say that all AI models are trending towards convergence in performance levels, which could render these models a commodity and instead shift the value creation towards apps built on top of these models. Interesting times lie ahead. #artificialintelligence #ai #aichatbots #chatgpt #googlebard #llms #google #openai

Artificial Intelligence is rapidly shaping our future, But where can we go to learn about it? These AI blogs are an excellent way to learn about artificial intelligence, even if you’re a beginner or are more progressed in your knowledge. Hundreds of AI blogs are available, but I picked only the best. So be sure to check out these top 6 AI blogs to jumpstart your knowledge in 2024: 1?? OpenAI Blog: Dive deep into the world of AI with updates on ChatGPT, OpenAI's groundbreaking tool for human-like conversations. From AI research to machine learning, this blog has it all! 2?? Machine Learning Mastery: @Dr. Jason Brownlee's blog is your go-to for mastering machine learning. Explore topics like deep learning, computer vision, and more to ace your AI skills. 3?? Amazon Web Services (AWS) Machine Learning Blog: Amazon's AI tools are revolutionizing businesses. Learn about machine learning frameworks, AI services, and cloud infrastructure for ML to solve real-world problems. 4?? Google AI Blog: Google's AI advancements are shaping the future. Stay updated on ML, NLP, computer vision, and ethical AI to witness innovation in action! 5?? There's An AI For That for That: Discover over 9000 AI tools for various tasks! Whether it's SEO, digital marketing, or cybersecurity, find the perfect AI solution swiftly. 6?? Massachusetts Institute of Technology AI Blog: Explore the forefront of AI research with insights from MIT's experts. From deep learning to robotics, stay ahead of the curve in AI innovations. This is TOP on my list as well as the tech reviews I subscribed to. Remember that learning about AI concepts and becoming good at AI takes time and effort. Just keep reading and learning, and you’ll learn, how to use AI in your daily live. What are your favourite resources for learning AI? comment to help everyone. #ai #techtalk #ArtificialIntelligence #OpenAI #amazonaws #google