A Supermarket AI Meal Planner App
A supermarket AI meal planner app suggests recipes that would literally kill people. Nvidia is now unstoppable, real or fake? Stay tuned as we unveil some of the biggest and most recent AI news that you’ve probably missed in the past weeks. Next up, we have Open AI facing financial struggles amid the crazy AI demand. So not everything is sunshine and roses in the world of artificial intelligence.
Extremely Popular AI Startup
Open AI, an extremely popular AI startup, has reportedly started sounding alarm bells about their financial sustainability. By now, you’re definitely familiar with Open AI’s headline AI products like Dolly for generating images from text prompts or chatGPT for language generation. The problem is, all of these complex AI systems require massive amounts of computational power and a staggering amount of money. Recent reports indicate that Chan GPT alone costs Open AI approximately seven hundred thousand dollars per day to operate. And that’s even with certain constraints still imposed to limit its usage. With no limits in place, the real ongoing cost of Chan GPT could be closer to three to four million dollars daily. So when you factor in all of the other services like Chad GPT and Dolly, Open AI’s total ongoing computer bill is estimated to be well over 100 million dollars per month, possibly much higher than that. Making matters worse, their crazy high demand from eager customers and users could start drying up over time as well. Without finding any new funding sources or revenue streams, Open AI could potentially face bankruptcy and shut down within the next couple of years. And let’s be real, that would be a catastrophe for the AI community, given Open AI’s groundbreaking research contributions. All we can do is hope that they can navigate these financial struggles successfully.
AI-Generated Videos
I want you to take a look at these two videos right here and try to guess which one was AI-generated. In the early days of computer programming, a significant historical anecdote is known as Grace Hopper and Debug. Both of them look convincing, but maybe the left one because the hand movement is kind of weird, or maybe the right one because of the lip movement and kind of the facial expressions. Well, to tell you the truth, both of them are AI-generated. Hey, John has recently made massive enhancements to their lifestyle avatars’ video quality and fine-tuned their voice technology so that it mimics any unique accent and speech patterns almost perfectly. This is a huge leap in AI generation.
The Grace Hopper Superchip
The GH200 degrees Hopper Superchip is ready to change the world. This powerful new chip can be used to train and deploy even more powerful AI models. It’s the world’s fastest AI accelerator for generative workloads. The Superchip is based on the ARM architecture, which is more efficient than the x86 architecture used in traditional servers. It is easily scalable to build data centers of any size. The Grace Hopper Superchip could be used to train large language models, generate realistic text, translate languages, and answer questions in a comprehensive and informative way. It could also be applied to healthcare, developing new medical imaging techniques that can diagnose diseases more accurately and quickly. In finance, it can create new financial models that can predict market trends more accurately. Some even think that it will make it possible to solve some of the world’s most challenging problems, such as climate change, disease, and poverty.
The GH2’s Performance
Nvidia displayed some impressive benchmarks comparing the GH200’s performance against their older A100 chip. For context, the A100 itself is already lightning quick, able to process requests on a giant 650 billion parameter NLP model at latencies of 50 milliseconds. But Nvidia showed the GH200 reducing that same inference latency down to just 30 milliseconds. So we could see double the throughput on enormous models with the new chip, with lower overall power consumption too.
Google AI’s Adaptive Ada Tape
Moving on to more technical AI news, researchers at Google have introduced an intriguing new technique called the Ada Tape. It makes neural networks more flexible and efficient by incorporating specialized tape tokens. The key innovation is adaptive tape tokens that are generated dynamically based on each problem. These tape tokens allow deep learning models to analyze and estimate the complexity of a task or problem that they are trying to solve. Then, based on that complexity estimate, the models allocate more or less computational resources accordingly. This translates into tangible benefits for users, optimizing compute usage and improving performance.
Microsoft 365 Integrations to Aid Frontline Workers
Microsoft announced some promising new capabilities coming to Microsoft 365 to better support frontline workers in various industries. Frontline workers often face challenges such as dispersed and remote work, limited access to key data sources, and handling sensitive information. To address these concerns, Microsoft is introducing new AI-powered features. These include a virtual assistant called Copilot that can handle common tasks automatically, personalized announcements for managers to broadcast important messages, and Windows 365 Frontline, a secure cloud-based version of Windows 10 purposefully built for frontline roles. These innovations aim to streamline workflows, close information gaps, and boost productivity and capabilities for frontline workers.
Thank you for watching this video. If you enjoyed it and would like to see more, feel free to check out our latest videos. And don’t forget to show some support by liking this video and subscribing so you don’t miss out on any future videos that we make.
