Terrence Hunter’s Post

The new toy is out: #GPT-4o "omni" from #OpenAI Took me few minutes to switch from my first thoughts: is just another gpt4 turbo version..where is the real news? I think it's in this part: "It can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in a conversation." This surely will enable more real time use cases across all industries! Now time to hands-on! https://lnkd.in/g3r5iCYT

This seems to be a better breakdown of the GPT-4o features vs. the demo earlier: https://lnkd.in/eFZUxHK5 Lots of really interesting stuff in here.

Well that's not what I was expecting on day one of Shipmas. OpenAI went straight in there with a new $200 per month option for unlimited access to their smartest model, OpenAI o1, as well as to o1-mini, GPT-4o, and Advanced Voice. It also includes o1 pro mode, a version of o1 that uses more compute to think harder and provide even better answers to the hardest problems. Have a look at the performance in the charts below. ChatGPT Pro is aimed at researchers, engineers, and other individuals who use research-grade intelligence daily. There are some cool new features, but if you're not in those categories, this subscription would be sledgehammer to crack a nut territory. Read more 👉  https://lnkd.in/edtvx3t9 Let's see what day 2 brings.... #AI #OpenAI #ArtificialIntelligence

During a flurry of other AI announcements this week, interesting to see OpenAI's #o1 come out today after o1-preview in Sept. Quite extensive safety testing results in the model card, but as expected no tech details on the models itself, and actually relatively sparse on accuracy. One thing to note is that o1 is more reliable, so they tested each test questions 4 times and required all attempts to be correct in the bottom graph. Only then pro-mode is showing a real edge. The models can still take tens of seconds or more to come back with a result. Let's see what else OpenAI has in store for their advents calendar, they promised news every day for the coming 12 days. My bets are at least on GPT4-o improvements, Sora and potentially also virtual world generation for training agents. Blog post https://lnkd.in/epCEdKaQ, model card https://lnkd.in/e3irkp7w, video https://lnkd.in/eeM29RTw

Our own ‘Jarvis’ ! I’m sure many of us wish we had Iron Man’s assistant Jarvis when we watched the movies. OpenAI is making this closer to reality now with the unveiling of GPT-4o. GPT-4 has been such a fantastic tool when it was launched about 1.5 years ago and now with the latest version, it’s now able to handle more capabilities across text, vision and audio (hence, the ‘o’ in the name for Omni ie. across all the channels). AI is progressing so rapidly that it’s both exciting and scary. Exciting because of the many possibilities it brings; scary because we’ve probably watched too many movies that has Skynet. In the video, GPT-4o can be seen singing a lullaby in whatever styles that you want and also analysing the environment you are in and then giving you comments. It’s even able to have 'emotions'. Wow ! With this progress, AI will probably drop the ‘A’ in the future and just be Intelligence ! #AI #OpenAI #GPT4o #Innovation #ArtificialIntelligence

Last week we released our first AI-enabled capabilities within Raygun! 🎉 This capability dramatically helps developers accelerate bug fixing and resolution. Here's how it works: 1. Your Raygun instrumented app sends any errors to the service 2. We group the errors by root issue, so you're focusing and able to prioritize bugs 3. When investigating a specific issue, AI Error Resolution will suggest fixes and code changes. You can chat with it for more information, and see conversations across your team. In our own use of this capability, we find it's a significant accelerant in the error resolution process. In many ways, I find it to be a much more satisfying version of Rubber Duck Debugging: https://lnkd.in/gYGynVY7 So jump on in if you're a customer, or start a trial if not, plug in your OpenAI key and away you go with crushing software bugs, and leveraging advanced AI to improve the productivity of your organization.

OpenAI has just launched GPT-o1, a new model redefining AI reasoning by significantly extending its thought process before delivering answers. Unlike its predecessors, which focused on pre-training, o1 invests in prolonged inference, sharpening its logical prowess. While it might not consistently outperform GPT-4o across every metric, GPT-o1 excels in tasks demanding deep reasoning. But there's more! Alongside this release are o1-preview and o1-mini: - o1-preview: A streamlined version perfect for refining specific use cases. - o1-mini: A cost-effective option delivering nearly the same performance as o1, especially impressive in STEM challenges. Both models are now accessible to ChatGPT Plus and Team users, with a broader rollout on the horizon. Currently, users have a limit of 30 messages per week on o1-preview and 50 per week on o1-mini. Looking Ahead: OpenAI envisions the o1 models achieving extended reasoning times—from seconds to potentially weeks! This leap could unlock advancements like drug discovery and theoretical mathematics. The future of AI reasoning is here, and it's poised to transform industries. Are you ready to explore what GPT-o1 can do? #OpenAI #GPTo1 #AIInnovation #ArtificialIntelligence #MachineLearning #DeepLearning #AIResearch https://openai.com/o1/

Just seen the new OpenAI GPT-4o features processing vision and audio in real time. Those examples are incredible - definitely check it out here https://lnkd.in/gGYBFwZg

The new OpenAI o1 Series has landed, but unlike the jump from GPT-3.5 to GPT-4, this one doesn’t feel like a game-changer. Yes, there are improvements, but nothing groundbreaking. For coding? Claude 3.5 Sonnet still takes the crown as the top performer. That said, the o1 models’ reasoning skills have definitely levelled up compared to GPT-4o, which hints at exciting future possibilities. But let's be clear—this is not GPT-5. It feels more like an extra Chain of Thought (CoT) layer on top of existing models. Exciting? Sure. Revolutionary? Not quite yet. #AI #OpenAI #ClaudeSonnet #GPT4O #ChatGPT-o1 #AIInnovation #TechBuzz #FutureOfAI

A New Chatbot called GPT2-ChatBot is Released,  but nobody knows where it is from! 🤖 With its apparent ability to match GPT-4's strength, the gpt2-chatbot has gained significant traction. This is the reason it's getting noticed: 1. Advanced Capabilities: It showcases advanced agentic skills surpassing those of GPT-4, breaking down complex tasks into manageable parts and effectively planning their execution, especially shining in coding tasks. 2. Community Buzz: Despite its mysterious origins, most likely to be from OpenAI, it has already caused a stir with its unconventional release. 🌐 3. Security Concerns: Within hours of its release, the gpt2-chatbot was 'JAILBREAKed'. The term 'JAILBREAK' refers to the technique used to bypass system restrictions, raising questions about safety and control. ⚠️ 4. It's taken down after 24 hours of public access, without any comments. 💬 What are your thoughts on the potential and risks of such advanced AI capabilities? You can read the full article about this new LLM player now 👇 https://lnkd.in/d7V_9Ddn #AI #LLM #MachineLearning #ArtificialIntelligence #OpenAI #GPT4