“Building Nvidia turned out to have been a million times harder than I expected... If we realized the pain and suffering and how vulnerable you’re going to feel, the challenges that you’re going to endure, the embarrassment and the shame and the list of all the things that go wrong, nobody in their right mind would do it.”
Last week, Google announced the launch of Gemini—a new family of AI tools that encapsulates the concept of Massive Multitask Language Understanding (MMLU). MMLU can be considered as a seamless integration of various Google AI libraries and tools capable of handling different types of content and media, such as text-based, images, videos, and audio files.
What are the implications of Gemini for the average person?
Gemini is set to provide more robust Google search results. For instance, a search for “British diplomat who referred to Netanyahu as a 24-carat bullisher” will yield results linking to YouTube video clips featuring the statement. Big Brother becomes increasingly powerful.
Very large language sequences models
The most potent Gemini language model (Ultra) was trained on datasets consisting of 32,768 (2^15) word (token) sequences. This stands as a testament to the achievements of the hardware engineers developing Tensor Processing Units (TPUs) at Google. The language model we presented for illustrative purposes considered average-length sentences containing 17 words. To put it in perspective, the official English translated version of Putin's famous 2021 essay (some would say rant) on Ukraine was just over 11K words.
No details or timeline on Gemini developer license
While Google's marketing material suggests that developers will have access to Gemini libraries, there are no details or timelines mentioned in the official documentation. The lightest version of Gemini (nano) is designed to operate on local devices.
Altman's ousting from OpenAI in the media spotlight
Several recent press reports have delved into the events leading to the ousting of Sam Altman from OpenAI. It took several years for OpenAI to achieve sustainable success. Until last year, the startup, founded in 2015, had fewer than 200 staff—hardly fitting the typical rapid growth startup story.
Most reports gloss over the blood, sweat, and tears behind every successful startup story. Generous donors who bankrolled the company with tens of millions of their own money for years against the odds. The first fifty employees who, for several years, turned down higher-paying jobs at Google or Facebook with better work-life balance because they believed in the greater purpose they were working towards. Young developers snorting Adderall to maintain a grueling work schedule of 17-18 hours a day, six days a week. Meanwhile, management teams strive to keep payrolls going to ensure employees stay motivated and focused on the mission.
Commercial realities must be reckoned with along the way. Something most people should be able to understand.
NVIDIA co-founder and CEO Jensen Huang says it like it is above… like the child birth pains which nobody ever talks about. Right now there seems to be too much concern for the health of the baby's father.
