Howdy from Raleigh,

Well, I’ve teased it long enough.

Time to tell you about my new job at a startup that’s trying to change the trajectory of AI.

Not all AI startups are new entrants.

Howso (formerly Diveplane) was co-founded in 2017 by Dr. Mike Capps (CEO), Dr. Chris Hazard (CTO), and Mike Resnick (Technical Fellow).

Capps is the former President of Epic Games (Fortnite, Unreal Engine, Gears of War). Hazard is a leading expert in ML with a focus on AI for trust and reputation. Resnick is a long-time collaborator of Hazard’s.

They had one mission: make trustworthy AI the global standard.

This stemmed from the co-founders’ shared concern that society would adopt black box AI to make critical decisions.

Spoiler: their concern was valid. Now, in 2023, black box AI has since taken off in the form of Alexa, Chat GPT, and basically every other form of mainstream AI.

What’s black box AI? It’s AI that can’t be explained or audited.

Basically, we put information into a “black box” or a neural network and get a result. We don’t understand how we arrived at said conclusion. We cannot trace how we got there. We trust it because…well - because.

In other words, we trust, but we have no way of verifying.

To the Howso team, that was unacceptable. Consider these human critical decisions that black box is currently influencing:

College admissions. Mortgage decisions. Parole hearings. Surgery approvals. Insurance claims.

If we’re going to integrate AI into our society (which we are - no going back on that), then we need a way to explain how we arrived at decisions that directly impact the well-being of humans.

That’s why Howso exists: to provide AI/ML practitioners with the power to understand and control the decisions they make.

That way, we can identify and mitigate potential bias, prevent hallucinations, and build greater understanding in society.

How did I get involved?

Back in January of this year, I had searched “ethical AI Raleigh-Durham” to try and find local community that was talking about the future of AI.

An article featuring a Raleigh-based startup called “Diveplane” popped up.

The tl;dr is that, as of yesterday, Diveplane has been renamed Howso. The rebrand was in-part why I was brought on - more on this below.

I read about them and found their work interesting, but I was heads down on trying to make Feather work. As a result, I didn’t do anything with it.

About 3 months after I had read that article, I attended Raleigh-Durham Startup Week. Chris Hazard (co-founder, CTO of what was at the time Diveplane) gave a talk at RDU Startup Week. I had the chance to see him speak, and I really enjoyed what he had to say about the dangers of black box AI.

Time passed - nothing crazy came of it because I was still fighting for Feather.

Finally, fast forward 2 months later to late May of this year. I was randomly reviewing the Duke Capital Partners website. They’re the venture capital arm of Duke University. I’m a Durham guy. They’re a Durham VC. Why not stay up to date on what they’re doing?

Lo and behold, I saw Diveplane listed as a portfolio company for Duke Capital Partners - small world. I clicked on their site, navigated to careers, and saw a listing for a “Product Marketing Manager” role.

At the time, it was becoming clear that Feather didn’t have a path towards becoming sustainable. I was in search of a role that allowed me to use my product and blogging experience to write for an AI startup.

This role met those requirements. I applied immediately and went to the address that they listed online to hand a cover letter in-person.

Bad news: it didn’t work. They had since moved out of that specific office.

Good news: that afternoon, while driving back home from my failed attempt to get a physical cover letter in their hands, the HR manager had emailed me requesting to speak.

That’s when I began the interview process.

Why did I join?

There is an exact moment when I knew I’d join (if they’d have me).

I sat down with Chris Hazard during the interview process. He is the CTO and person who I’d briefly met at RDU Startup Week. I asked him what success looked like over the next 2, 5, and 10 years for the company.

He put it something like this: mission zero is advance trustworthy AI in the world. We hope we’re the ones to do it. But, even if we don’t, we will be happy if someone can make it happen by building off of what we’ve done. Mission #1 is make enough money to hire great people, so we can advance mission zero. Mission #2 is to generate value for our shareholders. We operate zero, one, two - in that order.

That type of prioritization was wildly unlike anything I had ever been a part of. I wanted in. The timing, the role, and the work environment were all a fit, and I joined in early August.

One additional requirement that the role met: my need for in-person interaction. My first job out of college placed me on a remote project as my first client engagement. Six months later, the pandemic happened. As a result, I’ve never collaborated with a team in-person in my career.

Luckily, Howso just built a beautiful new office. It opened my first week of work. I now operate on a hybrid work schedule where I can maintain flexibility but also engage with folks in-person. It’s awesome.

What are my responsibilities?

I’ve now been serving in the product marketing role for ~6 weeks. My responsibilities are generally as follows: develop marketing collateral (blogs, pitches, demos), learn and map buyer personas, contribute to the product roadmap, and inform our go-to-market strategy.

My first big project? Write a website to help launch the new brand.

For more on the rebrand decision, you can read the executive team’s statement here.

They felt that Diveplane didn’t fully capture our mission to advance trustworthy AI as the global standard.

“How so?” is a question we ask when we seek clarity. We use the phrase to uncover how we arrived at a conclusion. It perfectly captures the intention behind the AI that we build.

So, what does Howso have to offer?

Currently, we have an ML framework (Howso Engine) and a synthetic data product (Howso Synthesizer).

Let’s start with Engine. It’s the basis of everything that we do. It allows you to generate output with exact attribution back to the input data, allowing for traceability and accountability of influence.

In other words, Engine is a fully explainable and auditable platform that empowers you to understand how AI makes decisions.

Then, there’s Synthesizer. It’s a product built on top of the Howso Engine. Folks need huge swaths of data to train machine learning models. The issue is that a great deal of their data cannot be used for this purpose due to privacy concerns. Plus, they don’t have enough data. That’s where synthetic data comes in.

Synthetic data is “twin” data derived from your original dataset without any of the personally identifiable information. You can even augment your data so that you have enough to train your model.

There are a number of synthetic data platforms on the market.

Other platforms use blackbox AI to generate synthetic data (meaning it’s not explainable). Synthesizer uses Engine (meaning you can fully understand how your data was generated). That’s the Howso difference.

With yesterday’s rebrand, we also launched our open source version of the Engine. We’re hoping to get it in front of data scientists across academia, research, and the Fortune 500 so that they can build explainable AI models using our tools.

Engine and Synthesizer are only the beginning, and I’m excited to be a part of positioning our products for success.

Here’s to a wild, winding journey in the AI space over the past 10 months and to making trustworthy AI the global standard - cheers.

What I’m paying attention to:

A work of art from my friend, Dylan. This is the energy that I’m taking into some PTO which starts today

Thanks for reading

Did you enjoy this post? If yes, then Howso? (See what I did there.)

Shoot me a reply and let me know,

Josh

Want more Build content? Check out the links below

Previous Newsletters // Essays // Twitter // LinkedIn