The path to AGI will take several decades as Andrew Ng stated yesterday.

In the meantime, it’s more practical to focus on developing AI agents that can function as effective human assistants making our lives better.

The Internet of Agents > The Internet of Things

Image taken from Davide's post on Agents
@davidecrapis

ethereum protocol research /ethrd + Crypto x AI /dacc 

AI Agents are ChatGPT on Steroids 💉

🔴 GPT-3.5 achieves 48% accuracy
🟠 GPT-4 achieves 67%
🟢 GPT-3.5 AI Agent achieves 95%

We often use LLMs like asking someone to write an entire essay in one go, without fixing any mistakes or making revisions.

However, with an agentic workflow, the LLM can work through a document in steps:
- Planning an outline
- Deciding if web searches are needed
- Writing a first draft
- Reviewing for weak points or unnecessary details
- Revising based on findings 

This approach lets the LLM refine its output multiple times, leading to better results than writing in a single pass.

The above is a partial summary of an article on Agentic Design Patterns by Google Brain founder Andrew Ng.

The intersection of AI and crypto is inevitable

Vader

@0xvader