February #TechBytes - 1st part

Here is a roundup of some of the tech news you may have missed recently selected by Etiqa!

Ours is a journey inside the latest advanced technologies, news about the most exciting startups and scaleups from around the world, and updates on specialized, high-risk industries.

OpenAI unveils Sora: AI-powered video creation tool

OpenAI introduced Sora, a groundbreaking tool that transforms textual input into captivating videos.

The name "Sora," meaning "sky" in Japanese, was chosen to evoke the concept of limitless creative potential. This AI tool promises to produce highly detailed, realistic, and high-quality videos, surpassing many similar tools currently available in the market. Sora-generated videos, lasting up to a minute, may feature multiple characters, diverse shots, and frames.

OpenAI showcased Sora's capabilities with a 17-second video set in snowy Tokyo, Japan, featuring two people strolling hand in hand amidst cherry blossoms, shops, and restaurants with Japanese signage.?

Currently accessible to a select group of designers, artists, and videomakers chosen by OpenAI, Sora is in a testing phase, with users providing valuable feedback on their experiences. Although not yet available to the public, OpenAI acknowledges that the tool has some limitations, particularly in representing complex scenes or displaying cause-and-effect relationships. For instance, if a character takes a bite of a cookie, the cookie remains intact in the subsequent part of the video.

OpenAI is committed to collaborating with regulators, artists, and educators to address concerns and prevent potential misuse of Sora. They recognize the challenges associated with predicting all possible use cases and potential abuses of their technology, emphasizing the importance of responsible AI development.

Nvidia's meteoric rise: now the fourth most valuable company worldwide

英伟达 , renowned for its chip production, has soared to a valuation of $1.83 trillion, narrowly overtaking Google's parent company Alphabet, according to Bloomberg. This positions Nvidia as the fourth most valuable company in the world, trailing behind Microsoft ($3.04 trillion), Apple ($2.84 trillion), and Saudi Aramco (approximately $2 trillion).

The unprecedented growth of Nvidia is attributed to the surging demand for artificial intelligence (AI). Currently, the company is at the forefront of AI development with its latest chip, the H100, powering a majority of large-scale language models in use today, including OpenAI's ChatGPT, as well as various AI projects by 微软 , Meta , and 亚马逊 . Nvidia's stock has surged by approximately 49% this year, adding a staggering $602 billion to its market value.

However, Nvidia faces stiff competition as major tech players like 英特尔 and AMD are engaged in a chip race, each aiming to create its own GPU to challenge Nvidia's monopoly. Undeterred, Nvidia, led by billionaire Jensen Huang, is on the verge of releasing an upgraded chip, the H200, featuring enhanced memory capacity and bandwidth compared to its predecessor. In early February, Reuters reported that Nvidia invested $30 billion in a dedicated division to assist other companies in creating their own chips, ensuring a potential gain even if competitors opt to build their own.

ChatGPT with memory feature: enhancing conversational intelligence

One of the key limitations of ChatGPT was its lack of memory. However, OpenAI has recently introduced a groundbreaking update – ChatGPT now has the ability to "remember" past interactions with individual users. This means you can instruct the chatbot to recall specific details for future conversations.

For instance, if you ask the program to compile a bullet-point list of the day's topics, it will seamlessly pick up where it left off, providing helpful suggestions based on prior interactions. To safeguard your data, you have the option to disable memory storage in ChatGPT settings, ensuring that the artificial intelligence doesn't utilize stored memories. You can also instruct the chatbot to forget previously remembered details, effectively resetting its knowledge for new instructions.

OpenAI has trained ChatGPT not to save conversations with medical professionals, particularly those involving sensitive information. For those seeking memory-free conversations, the option of temporary chat exists. These temporary chats won't appear in the history, won't use memory, and won't contribute to model training.

Developers creating custom chatbots, known as GPT, can also incorporate the memory option in their projects. However, they must provide a simple and quick way for users to enable or disable it based on their preferences.

European Office for Artificial Intelligence: a key player in AI regulation

Operating under the Directorate-General for Communication Networks, Content, and Technology, the European Office for Artificial Intelligence holds the responsibility of enhancing the Union's competencies in the field of AI and contributing to the centralized enforcement of AI regulations. Specifically, the AI Office will play a pivotal role in applying and overseeing the new regulations for general-purpose AI models. This includes the development of codes of conduct to detail the rules, classification of models with systemic risks, and monitoring effective implementation and compliance with the regulations.

The AI Office is empowered to request documentation, conduct model assessments, investigate reported incidents, and require providers to implement corrective measures. Furthermore, it will ensure the coordination of AI policy and collaboration among the Union's institutions, bodies, and agencies, as well as with experts and stakeholders. Establishing strong ties with the scientific community, the AI Office will support law enforcement application, serve as an international reference point for independent experts and expert organizations, and facilitate exchange and collaboration with similar institutions worldwide.