Found this doc from Google Cloud. I found this doc a good representation of a lot of advice you hear over the years regarding startups - these just have an AI focus.

General observations

More compute available combined with lower costs will make AI ubiquitous. I think this is right - the cheaper and easier a new paradigm (including embedded in different forms) is to run and access, the more likely it will be embedded everywhere.

A few people mentioned that adoption will be slower than expected, which also aligns with my experiences to date. People and organisations are resistant to change.

There were some mentions of AI with deep personal and emotional understanding that will make them essential parts of our lives. I find this a quite dark and dystopian, but I am making note of it to not lose sight of the possibility.

There is some commentary on progress slowing and focusing more on ROI of a product rather than a fancy demo. The time is now to seize the opportunities these tools have provided us.

Quotes

Software teams will be lean:

The future of software is doing more with less

• Chamath Palihapitiya

I'm not sure if this is cope, but this helps with my preferred end-state of all of this:

AI will continue to be a tool to augment human capabilities, not replace them. The concept of AI-based employees perpetuates a limited perspective that hinders the true potential of both AI and human intelligence.

• Arvind Jain

Stay focused - I am guilty of not doing this with my HR system:

Focus on building point features, not point products. Prioritize developing high-value, specific functionalities instead of creating comprehensive software. This allows competition in markets where larger companies could replicate features using AI.

• Chamath Palihapitiya

another one:

Instead of aiming for general-purpose AI, focus on solving specific niche problems with a lot of depth.

• Dylan Fox

Prioritize evaluations and continuous performance improvements of your system (through collection of high quality data):

Prioritize evaluations. Develop evaluations immediately to better scope the problem. Having clear metrics and ways to assess the performance of an AI system are crucial from the outset.

• Harrison Chase

Build the amazing idea right now that you think is currently too expensive with the expectation:

that compute will become faster, cheaper and more reliable as time goes on

• Amin Vahdat

No more prompting:

Design your user experience in a way that helps people accomplish their goals without having to use prompts.

• Raviraj Jain

The mother of all advice

"Move fast to market and talk to clients, iterate on what works and ditch what doesn't."

Classic. So true yet this is the hardest thing to do.