AI Newsletter: May Edition
Image generated by ChatGPT: Man reading a newspaper

AI Newsletter: May Edition

Welcome to the May edition of our AI Newsletter, a collection of the latest noteworthy developments and advancements in the rapidly evolving field of artificial intelligence. Here’s a quick overview of what we will cover:

  • The More You Use ChatGPT the Better It Gets
  • Meta Llama 3 is Finally Released
  • Adobe Announces New AI Features?
  • Heard of Poe?
  • The AI Race Between Google and Microsoft Continues
  • Leonardo AI Releases ‘Style Reference’?
  • The Robots are Here!

Of course, by the time you're reading this, there might already be an even bigger AI story making headlines - that's just the fast-paced nature of the field. Ideally, I'd put together a weekly newsletter to keep up with all the developments, but given time constraints, a monthly edition will have to suffice. So, let's dive in!

The More You Use ChatGPT the Better It Gets

This newly implemented memory feature in ChatGPT enables the LLM to retain information across user sessions to provide more tailored responses. OpenAI initially announced it in February 2024, but it has been rolled out gradually, only beginning to appear in user accounts in late April 2024.?

With this feature, ChatGPT can remember specific personal details about the user, in order to become more useful and personalised over time. Of course, this also raised privacy concerns, however, Open AI has added a feature for users to manage and delete memories, ensuring total control over their data.?

Meta Llama 3 is Finally Released

Meta, the company behind Facebook, has released Llama 3 - its new state-of-the-art AI model, which is being open-sourced. Whilst it is available in the US, it currently isn't available to UK users.?

Initially, Llama 3 comes in two versions: an 8-billion-parameter model and a 70-billion-parameter model. According to benchmarks, the 8-billion-parameter version performs on par with other open-source models like Claude and Gemini Pro. However, the highly anticipated 400-billion-parameter model, yet to be released, is projected to compete with closed models like GPT-4 and Claude 3 Opus.

In terms of uses, it operates very much like other AI tools, ChatGPT for example. However, it does boast better performance with functions like; actively browsing the internet for responses and the ability to change/edit images in real time via text.

The company is also adopting a conscientious strategy in the development of Llama 3, accompanied by the rollout of innovative tools such as Llama Guard 2, Code Shield, and CyberSec Eval 2, all aimed at enhancing trust and safety.

Meta intends to invest quite a lot in this model, increasing its spending forecast for the year to $35 billion-$40 billion, which was a significant uptick from the earlier projection of $30 billion-$37 billion.?

Adobe Announces New AI Features?

At the NAB (National Association of Broadcasters) conference, Adobe showcased impressive AI video editing tools that generated a lot of excitement. There were three primary AI adaptions featured including:??

  • Object removal and addition: exactly what it sounds like, the ability to easily add and remove objects in images and video with AI.?
  • AI video extension: this innovative tool allows users to extend video clips using AI to generate additional frames continuing the action based on the existing footage.?

These AI-powered video editing and generation capabilities represent a potential game-changer for video creators and editors by streamlining complex tasks within their familiar editing software. You can see the full interview below?

Adobe Talks About the Future of AI in Premiere Pro at NAB 2024

Heard of Poe?

Poe has introduced an innovative feature called Multi-bot Chat. This feature intelligently selects the most suitable model based on the nature of the query posed by the user. Users can also directly summon specific bots by mentioning them.??

The Multi-bot Chat interface offers a range of options, including Gemini 1.5 Pro, Mistol Large, and GPT-4, allowing users to switch between models based on their specific requirements. This development suggests that the future of large language models may involve specialised models that excel in distinct tasks such as coding, writing, or medical analysis.

Poe's initiative appears to be a proactive step towards this potential future, where users interact with a unified chatbot interface that intelligently selects the most appropriate model for each query.?

The AI Race Between Google and Microsoft Continues?

Google recently announced a significant investment in AI infrastructure, with plans to spend at least a hundred billion dollars over the next several years. This move mirrors Microsoft's collaboration with OpenAI to construct a hundred billion dollar data centre aimed at bolstering computing power and advancing towards AGI (Artificial General Intelligence).?

DeepMind's head confirmed Google's commitment to building a similarly sized infrastructure to enhance its AI capabilities. This competitive push underscores the race within the AI landscape, with both companies striving to lead the charge towards achieving AGI, though not explicitly stated, it's implied by the overall narrative of their actions.

Leonardo AI Releases ‘Style Reference’?

Have you ever crafted an AI-generated image with a distinctive style, only to struggle with reproducing it? Leonardo's latest innovation, the Style Reference tool, seeks to provide a solution to this challenge.?

This newly integrated feature allows users to upload an image and prompt the AI to generate another image imbued with a similar style. Not only does this tool promise to streamline the creation of consistent branding and marketing materials, but it also underscores Leonardo's position at the forefront of AI image generation technology.

The Robots are Here!

In the field of robotics, Boston Dynamics has recently introduced the Atlas 001, a robot that has captured widespread attention for its eerily impressive manoeuvres (see video below).?

The Atlas 001 is a fully electric, highly advanced model designed with a focus on agility and practical application in real-world scenarios.

?It is capable of navigating diverse terrains and performing complex movements, including flips and other dynamic manoeuvres.

Equipped with sophisticated sensors and advanced control systems, the Atlas 001 can maintain balance, accurately perceive its surroundings, and interact effectively with various objects. Essentially, the future of robotics is here, and it is already in motion!

Key takeaways

As we conclude this edition of our AI Newsletter, it's evident (but not surprising) that the field of artificial intelligence is progressing at an unprecedented pace.?

The advancements we've covered, from ChatGPT's memory feature and Meta's Llama 3 release to the introduction of revolutionary video editing tools and the eerie Atlas 001 robot, are mere glimpses into the vast potential that AI holds.

Stay tuned as we continue to bring you the latest developments, groundbreaking discoveries, and thought-provoking discussions surrounding this transformative technology next month.

To learn more about AI visit our website and explore this new technology.

要查看或添加评论,请登录

社区洞察

其他会员也浏览了