Beyond A Billion Tokens, Midjourney Release Roadmap & Prompt Battle Royale

Beyond the Billion Tokens: Unpacking the Implications of Microsoft's LongNet

On July 19th, Microsoft unveiled a proposal for a transformer model that could theoretically handle up to a billion tokens, a development that could potentially revolutionize the field of large language models (LLMs). But as we celebrate this technological leap, it's crucial to also consider the broader implications of such advancements.

LLMs are deep learning models trained on a vast corpus of text data from the internet. They create a complex map of language, generating output based on input text. It's akin to having a sophisticated language GPS. But the journey to this point has been fraught with challenges, not least of which is the issue of context length restriction.

In simple terms, context length restriction is like trying to understand a conversation after walking in halfway. You can grasp bits and pieces, but without the full context, your understanding is limited. Microsoft's LongNet aims to solve this problem, theoretically capable of handling a billion tokens, thereby removing this major obstacle.

This development is undoubtedly impressive. It's akin to walking into a party late but still being able to catch up on every conversation. But as we marvel at this feat, we must also ask: What are the broader implications of such a development?

Firstly, the ethical considerations. As LLMs become more sophisticated, the potential for misuse grows. These models could be used to generate deepfake text, spread misinformation, or even conduct sophisticated phishing attacks. How will we regulate the use of such technology to prevent misuse?

Secondly, the societal implications. As these models become more integrated into our daily lives, they could fundamentally change how we interact with technology. But at what cost? Will we become overly reliant on these models? And what happens when they make mistakes, as they inevitably will?

Finally, the economic implications. As these models become more powerful, they could potentially replace jobs, particularly in fields like customer service or content creation. How will our economy adapt to these changes? And how will we ensure that the benefits of this technology are equitably distributed?

In conclusion, while Microsoft's LongNet represents a significant step forward in the development of LLMs, it also raises a host of questions that we, as a society, need to grapple with. As we continue to push the boundaries of what's possible with technology, we must also ensure that we're considering the broader implications of these advancements. After all, technology doesn't exist in a vacuum. It shapes our society, and in turn, is shaped by it.

Midjourney Releases Roadmap

Here’s a quick rundown of the most recent Midjourney Weekly Office Hours:

(1) Midjourney V5.3

• The V5.3 release has been delayed due to UI issues and compatibility with the mobile app.

• Working on in-painting and improving the styles.

(2) Midjourney V6

• At the moment, the V6 feels more like a V5.5.

• Its development will be continued than releasing it prematurely.

(3) Future projects

• The current priority is to build standalone web and mobile platforms.

• Several projects involving 3D, motion, real-time, and a secret storytelling research project are ongoing.

• The long-term goal is to offer video and 3D without raising prices.

• A potential new feature on generating images faster and in larger batches.

• Another new feature related to fast image generation is being considered for users who want to iterate quickly with fewer images.

• Increased server investment to avoid potential GPU shortages.

(4) Community updates

• Changes were made to the ranked images system in preparation for potential data collection.

• There will be more voice events, such as art classes, guest appearances, and prompt battles.

• The overhaul of the image archive downloads is ongoing.