michael's update #10
fall 2023
hello!
here's a seasonal update with three things:
what’s cooking: built an ai assistant coach for writing = finalist for open ai tooling competition
building community: mit tough tech summit
miscellany: 2023 science and tech news recaps in lab automation, ai x robotics, and synbio
my goal is to share important updates from the last few months. i like a menu format so if you’re curious for more, then i linked other things to explore.
enjoy,
michael
🌊
1. what’s cooking: built an ai assistant coach for writing = finalist for open ai tooling competition
for context, we’ve been exploring new program, community, and product offerings for what’s next for me. my goal is to focus on higher education market, and these last few months we’ve been exploring what role new machine learning models will play in personalized education experiences.
rather than focus on a new program pilot, we’ve been experimenting with prototyping an ai assistants for coaching. the first one is a writing coach. it works well enough to give structured feedback on writing and is trained to help with project based work.
students i’ve worked with over the last 2 years says it’s “8/10” and “similar to feedback you would give.” i’m suprised that even a rough proof of concept works this well. about 2/3 of the feedback has been accepted as useful.
we’ve used this to apply to different education grants to see if this can kickoff a larger build. we’re finalists for second phase of open ai’s future of education tooling competition.
the main learnings from this season’s focus:
ai coaching by itself is not enough, it does not solve for other factors like individual motivation or supportive environment. it’s one piece of a larger puzzle.
open ai’s new tooling will make personal gpts ubiquitous, meaning prestige and brand will win out in the first wave of education + ai tooling. people would rather work with the harvard ai writing coach than a random branded one.
a focus on product strategy is not enough, as anyone can make gpts now, so the barrier to entry is lower.
next up, there are other puzzle pieces we need to get right including channels for distribution and business model for intended use case.
more coming soon as product is not fully released.
bonus: for fun we hosted this 1 day virtual climate tech summit for the community. recaps here and here.
2. building community: mit tough tech summit
one highlight for helping build the tough tech community was helping at the mit engine’s annual summit. with a team of tks alum, we helped with ops for the conference.
they doubled attendance this year from 300 to 600+ attendees. one of my favoriate parts was the keynote from ceo of common wealth fusion systems, bob mumgaard.
top takeaways:
build an ecosystem from day 1
clear vision = motivation
have a 10 year master plan
break out of simple and build what it needs to be to work
when doing something new, there is no pattern recognition
bonus: a blog post reflecting on the social physics of unconferences and how they make better experiences.
3. miscellany: 2023 science and tech news recaps in lab automation, ai x robotics, and synbio
here are some research rabbit holes that reflected on interesting advancements in 2023 and exploring what that could mean for 2024, especially in the ai and hardware space.
always cooking something,
michael
ps. if you made it this far, you know what's up: recipe inspiration for the season. this shaved brussels salad with pomegranate was a fun one.