Claude 3: Possible AGI

Author

The Renaissance Times
March 06, 2024

AI self portrait, 2024

Welcome Artisan,

When our descendants look back at the 21st century, they will discover in their history books that the day humans invented / discovered AGI was one of the most important of the century, maybe ever.

Yesterday was maybe, possibly, probably not but there’s a chance, that day.

Let’s dive in.

Hot Off The Press

The One Big Thing

We may be reaching a point of maximum acceleration…

AGI stands for artificial general intelligence. When most people talk about AI and the crazy outcomes that could stem from it, they are usually talking about AGI or ASI (artificial superintelligence).

The exact meanings of these terms have now gotten muddled, but let’s define AGI as “a form of AI that possesses the ability to understand, learn and apply knowledge across a wide range of tasks and domains," per Gartner.

Modern LLMs are not AGI precisely because they do not understand, reason or apply knowledge across a wide range of tasks to achieve any outcomes. The way LLMs work is they take a massive set of initial training data, use it to create a model of outputs, and make predictions of each ensuing character based on this model.

AGI is the holy grail when it comes to this stuff. AGI is when a model is able to improve itself, thus leading to ASI.

Now that we have the terms out of the way, let’s talk about the big news of the day: Claude 3. We covered the release of these new models yesterday, but 24 hours in we are starting to see some crazy results.

It seems to be reasoning out math problems (which are typically very difficult for an LLM, as you may have seen with ChatGPT or other models when they hallucinate).

Claude also handles language translation well, and does so in a manner that suggests deep understanding of the language structure itself.

The crazy part about all of this is that Anthropic (the creators of Claude) recently revealed that during testing for Claude 3 Opus, the strongest model released yet, Claude actually suspected that the human performing the test was performing a test!

Read the following excerpt and try to see if you’ve ever witnessed a machine behave this way. I have not.

At this stage there is not much more to say but, wow! We have a new LLM that can handle math, language, self portraits, and is even beginning to show signs of self awareness.

If you have not played with Claude 3 I highly encourage you to do so. We may be witnessing the great acceleration in real time.

Presented without comment

How does an AI see itself?

Tools

Must have tools for every Renaissance creator to add to their toolkit:

  • Haiper: video-generation tool with its own AI model underneath

  • Rufus: AI powered chatbot by Amazon

  • SDXL Lightning: A fast text-to-image generation model that produces high-quality images

  • Attention: A tool to automate sales process, CRM updates, call coaching and analytical insights.

  • GuestLab: A tool for podcast pre-interview research

Deep Tech

The newest and coolest in the research world that you need to know about:

  • GDDR7 is the first JEDEC standard DRAM to use the Pulse Amplitude Modulation (PAM) interface for high frequency operations

  • Microsoft’s new Orca-Math AI outperforms models 10x larger

  • UDOP: A vision encoder, text decoder Transformer similar to Donut

  • Nomic AI new release: instantly search, download, and privately chat with any model on hugging face

  • A Safe Harbor for Independent AI Evaluation

Closing Thought

❝

Would an AGI tell us it was an AGI?

Canceling weekend plans to play with Claude 3

Work With Us!

The AI Renaissance is coming and we are building the best community of the people making it happen.

Contact us to sponsor your product or brand and reach the exact audience for your needs across our newsletter and podcast network.