Denver hosts Denver Startup Week (DSW) once a year and I was able to attend a couple sessions in September. A few folks asked me how the event went and here are my not-so-deep thoughts and notes.
Main takeaway: You shoulda been there.
And if that reminds you of Jack Handy's Deep Thoughts that's okay with me. š
DSW is a community event. Thanks to the sponsors, it's free for attendees. DSW is people powered, so they need you.
Attend the sessions, chat with others, meet and network with folks you don't normally see day to day.
If you're up for it, submit a talk and present. (I've presented in the past and am happy to help you brainstorm ideas or refine your talk.) Or volunteer and help organize. Reach out to folks like Lacey and other DSW folks and find out how you can help.
DSW is better with you involved.
Notes
Here are notes from a few sessions I attended (in chronological order). It was great to meet a few folks for the first time and hang out downtown. Not every session was recorded, so check YouTube for those that were.
Neurodiversity in Entrepreneurial Endeavors
The first session I went to was a panel on Neurodiversity.
My main takeaways:
- I appreciate the panelists vulnerability and honesty in sharing their personal challenges
- Try things and get help! It's okay if something doesn't work for you. Learn from each other and how others are coping.
- Give yourself grace
- Celebrate differences - and seek to understand what they are going through
Panelists:
- Sara Bates
- Tiffany Feingold (Guiding Bright Minds)
- Stephanie Cohen
- Josh Schuler (MeFlow)
Host:
Ibotta - Decoding AI
Ibotta folks led a talk on using AI/ML to drive product growth. Unfortunately, the talk focused on a vague "educational product" versus a core-in-house challenge at Ibotta, but was still good to learn from their experiences.
Speaker list is in the link above.
My main takeaways:
- Models and machine learning are only as good as the data they use (I heard this more than a few times during DSW)
- If starting with ML, try batch predictions or model output instead of real time output
- Include a model version column in the results (A neat idea that allows them to switch/test models behind the scenes)
And my favorite note:
If it doesnāt have automation, then it should not be in production
A really good point about code, whether AI/ML or not. Getting automation is place will pay off hundreds of times as you build and deploy your application going forward.
Keynote: Women of DSW | Build the Damn Thing with Kathryn Finney
Keynotes at DSW are a lot of fun - tons of excitement and energy in the room. This Q&A and with Kathryn Finney was really cool. She gave out signed books too.
It's on Youtube, so check it out.
Negotiating with OpenAI
This was a fun talk on building an app with OpenAI. The premise was "negotiation practice" and the app lets you negotiate a car sale with an AI bot. Fun idea!
The room was packed for this one (as you might expect). Great example of using good story telling to share knowledge.
Thanks Tyson!
My main takeaways:
- "LLMs are a Swiss Army knifeā¦but what we might need is scissors"
- Embedding encodes the semantic information in a block of text and can be used to search (vector similarity)
Check out the GitHub project for the app
Spoiler alert: Tyson said the bot used to give away the price of the used car in negotiations until they came up with the right prompt. :money-with-wings:
Data Privacy & Security Related AI issues for the Early Stage Company
This talk was a panel on general privacy and data concerns at early stage companies, where you may not have a dedicated security/legal/trust and safety team. The panel list is in the link above.
Panels are often wide ranging, so gonna just dump my notes here. My main takeaway:
- Security and data privacy issues affects M&A and deal price
You don't need every scenario covered, but you should have enough of a data security plan to provide answers and comfort.
Raw Notes
Need to have enough info to provide comfort
Instill discipline - where is data stored, who has access? Easier to do up front than in the middle of due diligence.
Start good habits early.
Weigh impact on engineers and product
Uplight - created an AI policy for engineers
Guidance document that explains why
Ask ChatGPT why you shouldnātā¦
Mockeroo/Tonic - create mock data sets
Levels of data protection
Confidential - anything not on internet
Sensitive - info that is within company that not everyone has
Can I use ChatGPT?
Check all your terms of use, agreements. Make sure not violating existing terms.
Check for data addendums with providers
How to claw back data once it leaves?
Even if you canāt execute due to resources and budget, produce something tangible (a doc outlining what you decided or would do later on)
Ask people that are leaving - they have time on their hands. 95% of what you need is in Google or MS suite.
Donāt copy and paste terms - other companies may not have legal counsel themselves.
Donāt back yourself into a corner in your first privacy policy.
Building an AI Empowered Workforce
JP had a great overview of how to weave AI into your organization and day to day. He highlighted the potential "mass disruption" of AI tooling and how we can align our teams with what AI does well.
My main takeaways:
- Leads should communicate: What are short and long term goals of using AI? Why should anyone care?
- Ensure safety and solving real world problems
- Do an internal audit. See what people need before throwing AI tooling at them.
More raw notes here
Experiment!
Make lightweight tools to solve real problems
3 camps - super fans, skeptical, apathetic
The hardware and investment is there - likely here to stay.
How to align with what AI can do well?
Do an internal audit. See what people need.
Group tasks into now / later / never
Are there tools that solve safely and effectively?
How does your org need to change?
Why is (a current problem) the case?
What could it look like?
How will you do it?
Top comments (0)