Developers of ‘Stuff and Things’
AIM Events
Hosting the World’s Most Impactful AI Conferences & Events. For Brand collaborations write to [email protected]
At the Microsoft Build 2023 conference held in May, the tech giant announced Microsoft 365 Copilot and coined the phrase “everyone is a developer now”. Copilot was built on the hype of ChatGPT and large language models, so, after ChatGPT, even a layman has legit become a developer.
And when everyone turns into a developer, the developer turns into a full-stack developer! That’s how the chatbot has transformed the app/website building ecosystem. Despite the hype around the “everyone’s a developer now” idea, there’s little data to back it up. But the opportunities that the GPT-4-powered chatbot has created for developers can be ostensibly measured.
In the website building world, the role has always been divided between backend and frontend developers, who needed specialised training to do each of these jobs. Now, ChatGPT has blurred the line between the two. Developers are now upskilled with “promptgramming”, who can single-handedly build the backend and frontend.?
Now, what will full stack developers do? According to tech experts, full stack developers, who are already proficient with a language, should aim to become full stack polyglot developers, with knowledge of different languages.?
A funny Twitter thread discusses that after the new ChatGPT development, he has 5+ years of experience with stuff and things. A user commented, “Awesome, I’ve got way more experience than that with Stuff and Things, including integrating them into such and such.”
Read the full story?here.
OpenAI’s Arch Nemesis?
Reka is out of stealth mode. Founded by former Google, DeepMind, Baidu, Meta, and Microsoft researchers, the company announced its Series A funding of $58 million, which was led by DST Global Partners and Radical Ventures, along with Snowflake Ventures. The company’s only product, which is still in beta testing, is Yasa, a multimodal AI assistant for images, videos, and understanding tabular data, something very similar to what OpenAI is building with GPT-4.?
The AI company aims to build generative models and push the frontiers in AI research, which includes developing a "universal intelligence”, referring to versatile AI agents capable of handling multiple modes of communication and languages. These agents are designed to be self-improving AI models and are specifically tailored for enterprise software applications.
Read the full story?here.
领英推荐
Deepfake Antidote Found
Deepfake image identification solution is here. Shivansh Mundram, an alumnus of IIT-Kharagpur and a LinkedIn employee, along with researchers from the University of California and Berkeley, found specific patterns in fake images generated by GAN (generative adversarial network). They created a simple linear model to detect such fake images. They used light-weight, low-dimensional models with relatively minimal training data that mathematically distinguished the definite pattern of StyleGAN faces from real profile images.
Mundram and his team are proud of this achievement but feel they have a long way to go. Though this method does work for GAN images, in other models like Stable Diffusion, which have different types of structural patterns built into them, they are still not able to detect fake images with these simple linear models. “But there are other methods we are working on which we’ll put out soon,” Shivansh explained.
Read the full story?here.
NVIDIA's GPU New Record
NVIDIA's flagship H100 chip has once again outperformed its competitors, achieving the best performance to date on a series of MLPerf training benchmarks. In collaboration with CoreWeave and Inflection AI, NVIDIA conducted tests using a cluster of 3,584 H100 GPUs hosted on CoreWeave's platform. These GPUs were interconnected with InfiniBand, enabling outstanding performance at both individual and scalable levels.?
MLPerf benchmarks measure hardware capabilities by evaluating the time required to complete specific workloads, including LLMs, computer vision models, CNNs, and RNNs. NVIDIA's H100 chip demonstrated remarkable performance across all these benchmarks, setting new records in the process.
Read the full story?here.