Zumper’s Post

Following the release of GPTs to the public by OpenAI today, I decided to take a different path and show you how I built my own system from the ground up fully locally on my PC! On the left of this screen, you'll see one of my custom-built GPTs tools in action. It's fully customizable, allowing me to tailor it to my specific needs. And the best of it — I wasn't limited to just one model! By exploring different options and comparing them to the top models like ChatGPT 4o, I was able to discover interesting variations that improved my results. The right side of this screen showcases my paid account for OpenAI's ChatGPT 4o, the current industry leader. I asked both systems one and the same question: "Can I plug in ChatGPT or other large language model as copilot in MS SQL Studio?" As you can see, the suggestions provided by my custom-built GPTs are similar, if not identical, to those offered by the top models. And I plan to take it a step further by incorporating internet search results to keep my models up-to-date and enhance their responses even more! Don't be afraid to explore different avenues and tools to find the best solution for you! Design-Experiment-Learn. Repeat 🔁 #AI #CustomSolution #TechInnovation #Streamlit #Flowise #Ollama #TestExperimentLearn #OpenSource

Similar to OpenAI's custom GPT Builder and Hugging Chat Assistants, a user can type in text what they want their bot to do. https://lnkd.in/g34FpjV4

A basic reasoning test between the free versions of OpenAI's ChatGPT-3.5 and Google's Gemini. Here's some differences in a nutshell: ⚡️ Gemini reasons, ChatGPT-3.5 doesn't (see image for comparison). ⚡️ Compared to GPT-3.5's text-only inputs, Gemini has option for audio + image inputs ⚡️ More speedy outputs in Gemini (more tokens/sec) ⚡️ Better context window in Gemini (32k) vs ChatGPT-3.5 (4k) Which one are you using? --- Are you technical? 👉 Join 200+ others in reading how large distributed ML systems are built: codecast.substack.com

PSA: For those with subscription to Perplexity and OpenAI. Perplexity recently released a new Sonar model available in their pro plan (LLama3 with their own secret sauce). Amazing speed and research. I will typically use the output I get from this and then use it as context with ChatGPT 4 or feed it back to Perplexity with the Claude Opus Anthropic (or Claude.AI) model to get amazing reasoning and strategic output. What's next? still trying to get a feel for Google Gemini... I might try it in the same way I tried Claude.AI models... but looking forward to mid-May to see how things change with pending OpenAI and Google announcements.

Thoughts on OpenAI Spring Update: - They should have let ChatGPT finish that story with Byte. They kept interrupting her. - GPT-4o looks interesting. This just proves that founders need to stop obsessing with naming things. The bar is pretty low! - Strong focus on speed and efficiency by the OpenAI team with this release. - The audio feature is damn cool. - The desktop app can read your screen and explain whatever is going on.

Some interesting insights from OpenAI’s Ask Me Anything session on Reddit. - Upcoming improvements to o-series models, including multimodality and tool use - Expectation of continued 10x decrease in inference costs - Emphasis on AI's potential in software engineering - Convergence of GPT and o-series models - Focus on autonomous task performance and human interaction in 2025 - Upcoming releases, not labeled as GPT-5 - Future capabilities in dynamic web page rendering for search queries - Imminent release of full o1 with noticeable improvements - Upcoming GPT4o for image and 3D content generation - Agents as the next major breakthrough - Use of citing and reinforcement learning to address hallucinations - Recommendation to build for an agentic future with cross-application workflow integration https://lnkd.in/eTJQpEM3

Hey, I rarely plug my newsletter here, but while everyone is doing boring demos of ChatGPT Search, we need new angles on what's at play: It's not OpenAI versus Google and Perplexity; it's a market share battle with Anthropic. With Search, OpenAI reinforces its position as a leader in the B2C market. They need it; Anthropic still has the best model with Claude 3.5 Sonnet and is growing fast as an API provider: • Claude 3.5 Sonnet is the most used LLM on Open Router • The latest version of Claude Sonnet broke the 50% SWE-verified benchmark • They're about to close a fundraising round with a $40B valuation On the other side: • No GPT-5 this year • New DALLE coming, but no date. My bet: Flux, Ideogram and Recraft cut their roadmap; when you're called OpenAI, you want to release a SOTA model, nothing else • Unstable teams with key people leaving the company So what's your bet? OpenAI or Anthropic? Put your vote in the comments 👇 Read and Subscribe for sources: https://lnkd.in/eWW-i2k6

🚀 Big news, folks! Tomorrow at 12 PM EST, OpenAI is set to unveil some fresh updates for ChatGPT and GPT-4. 🤖✨ Sam Altman spilled the beans - they've been hustling hard on something over at OpenAI. It's not GPT-5 or a new search engine, but it’s bound to be a game-changer! Stay tuned: [openai.com](https://openai.com/) 🌐💡 #TechUpdate #AIRevolution

ChatGPT Takes on Search (and Google) with SearchGPT! 🌐💥 Discover why the new SearchGPT from OpenAI is causing a stir and whether it can truly provide accurate information in a way that differentiates itself from legacy search like Google. #aiinsidepodcast #aiinside #ChatGPTSearch #AIInnovation #GenerativeAI #SearchDebate #TechNews #GoogleComparison #OpenAIvsGoogle #AIAdvancements #TechDiscussion #InnovationInsights

OpenAI just launched a mini version of GPT-4o. It can handle content up to 2,500 pages, including input and output. But it only costs less than a dollars to use. This smaller model follows the trend of other companies releasing compact versions. OpenAI says its performance is a bit lower than GPT-4.5 but better than GPT-3 Turbo. Switch to the mini model if you use ChatGPT API heavily. Did i saved your cost? Follow me!