OpenAI Unveils Sora, Google Intros Gemini 1.5, ChatGPT's New Memory, & OpenAI Set Sights on Google Search
OpenAI unveiled a new tool for AI generated videos, Google introduced an upgraded Gemini, ChatGPT now has a memory, and OpenAI may be going after Google search.
OpenAI Unveils Sora: A Leap Towards Photorealistic AI-Generated Videos
OpenAI has introduced Sora, a groundbreaking app capable of turning text prompts into photorealistic videos, marking an impressive advancement in generative AI technology.
Unlike its predecessors, Sora stands out for its ability to create videos with striking photorealism and longer durations, offering clips up to one minute long, a feature not commonly found in other text-to-video models. The technology leverages a diffusion model similar to Dalle-3's image generator and GPT-4's transformer-based engine, demonstrating how they can fulfill prompt-based requests while intuitively grasping cinematic grammar and storytelling nuances.
Despite its current limitation to select creators and safety experts for vulnerability testing, Sora has already showcased its potential through examples like the above California gold rush scene, demonstrating their future impact on both professional and social media video production.
Sam Altman took live requests on Thursday, showcasing how powerful Sora is. Sam responded with this video request of "Two golden retrievers podcasting on top of a mountain" in less than an hour.
Google's Upgraded Gemini 1.5 Has a 1 Million Token Context Window
Just two months after debuting Gemini, Google introduced Gemini 1.5.
Gemini 1.5 has significant enhancements, notably achieving parity with the recently introduced Gemini Ultra and outperforming Gemini Pro 1.0 in 87% of benchmarks
A groundbreaking feature of Gemini 1.5 is their massive context window, capable of processing up to 1 million tokens, dwarfing the capacity of previous models and enabling the handling of vast amounts of data simultaneously.
As impressive as Gemini 1.5 is, this announcement comes when OpenAI made three major announcements over the same week. Google may need to move faster, and we'll see how much they can catch up later this year.
ChatGPT Rolls Out Memory Feature
OpenAI announced a new memory feature for ChatGPT, initially available to a select group of free and Plus users, designed to enhance user interaction by remembering details from conversations.
This capability helps users avoid repetition, making their ChatGPT interactions more efficient and personalized. Users can control this feature, deciding what ChatGPT remembers or forgets, and even turn off memory entirely for complete privacy.
This update also extends to GPTs, allowing them to retain information for more tailored assistance across various applications.
As interesting as this sounds, we're curious how well this work in practice. Given complaints about ChatGPT laziness, which a bloated system prompt could be contributing to, this new feature could make ChatGPT even more bloated than it already is.
OpenAI Sets Sights on Google Search
As if we haven't talked about OpenAI enough, they're reportedly challenging Google with a new search tool, according to The Information. This move comes a year after Microsoft's CEO, Satya Nadella, intensified competition by incorporating Copilot AI into Bing, aiming to shake up the status quo in search engine technology.
With Google's Bard and Gemini, along with emerging players like Perplexity entering the fray, the competition among AI-powered search tools is becoming increasingly fierce, signaling a significant transformation in how users navigate the internet.