The Age of AI Podcast

Have you ever had difficulty understanding what someone was saying because of their accent?

Do you have an accent that others don't understand, or hold biases against?

As a speaker of many foreign languages, I've always found it to be a very fascinating process. Ask any foreigner in your country about their relationship with the local language, and you'll realise how the topic is packed with various emotions.

But as it turns out, AI can help us communicate perfectly with each other in spite of thick accents.

My guest Ganna Tymko is the CEO of SaySo, which uses AI (specifically deep learning) to transform heavily accented speech into standard English (i.e an accent from Britain, America, etc) — in REAL TIME, as you speak!

Just imagine how it would transform cross-cultural and cross-border communications! We dive into the technology and business aspects:

00:23 - Why Aman is getting desensitized to accents

04:53 - Why we struggle with accents

13:59 - AI projects: research VS design

23:13 - How Sayso's tech works

33:23 - How our mouth muscles get trained to produce accents

41:48 - Taking accent transformation tech to the market

47:06 - When training ML models for humans, use human metrics

51:05 - Closing

"Tech Fluent CEO" for non-technical entrepreneurs: https://aman-agarwal.com/tfc/

A few years ago, DeepFake technology appeared on the horizon.

It allows you to take  a person's video and voice recordings, and create a digital "avatar" that you can use forever. We saw synthetic videos of political leaders and celebrities saying things they had never said in real life, but which looked so real they were indistinguishable from the real thing.

While many were concerned, others saw a more ethical potential –making unlimited videos with "AI models," by simply giving them a script to recite.

One such company is Colossyan, and in this episode I sit down with their CEO Kristof Szabo to discuss how they're dealing with the technical and non-technical challenges of building such a unique, AI-first product.

This is an example of a product that would be unimaginable a few years ago, but which will be virtually everywhere in the future. AI celebrities (which do not exist) are already taking over social media and have scores of ardent fans.

Personally, I can imagine a world where 90% of actors in movies are not real human beings, and where almost every movie you see is "animated." Not sure if that's what will happen, but it's a definite possibility.

Here's what we discussed:

2:01 – Who are early adopters for a technology like this?

7:16 – Challenges in building the technology; what makes it so hard

13:08 – Why the "secret sauce" matters less than the product roadmap

18:00 – How to recreate body language for an AI avatar? (Also, Aman makes a non-European joke)

22:30 – Colossyan's origin story

27:00 – Lessons from talking to customers

31:36 – Fundraising cycle and product-led growth

37:00 – How to keep research and product teams in harmony

42:23 – Crazy idea: doing away with the "ML team" in an AI company?

46:48 – (Allegedly true) story about Kayak

(Ethics Policy: These opinions are 100% my own as an independent observer and educator. I don't own stock in guests' companies or their competitors, nor do I get paid by them in any form for any reason at the time of publishing, unless specifically stated. Episodes are also not intended to be an automatic endorsement of any company or its products and services.)

You know that the future of transportation is electric, and that future is already here, but it's not "evenly distributed" – some countries, companies, and people are making faster leaps towards switching away from gasoline than others.

For companies that operate a lot of vehicles (in transportation, municipal amenities, retail etc), it turns out that switching to electric vehicles is more complicated than it sounds. Which cars do you switch first, which drivers do you allow to drive them, etc.

It's a niche problem, but it's growing, and today I speak to an entrepreneur whose company is dedicated to solving it – and their solution uses AI!

My guest was Mateusz "Mat" Maj, the CEO of VivaDrive, and in this episode we chat candidly about all the different aspects of managing fleets of vehicles.

1:26 - What are green fleets?
2:30 - Why European companies care about having green fleets
4:45 - Building a "digital twin" of a company
11:20 - Using an AI system to improve car fleet operations (in-depth reveal)
20:30 - What an "AI fleet manager" does: big data vs ML
25:00 - Using ML to upgrade to green cars
29:20 - Why is it complicated to switch to electric cars?
38:00 - European power grids aren't ready for electric cars
41:00 - How VivaDrive approached their go-to-market strategy
48:10 - Fundraising strategy (and how they chose it)

The manufacturing industry is as diverse as the number of products in the market.

There’s always a never-ending journey to do things better. And one of the key "things" is quality control and safety-critical inspections.

Inspections are quite labour-intense and expensive, which makes them very unscalable. The way companies have gotten around this so far is by sampling a tiny few of the products and look for defects, and then extrapolate from there.

But with AI, there's a better way. Humans don't scale, but AI does.

Which brings me to my guest Christian Els, the CEO of Sentin AI, which provides software for AI-powered inspections in manufacturing.

But manufacturing is often considered one of the most conservative, slow-moving industries (factories are expensive and don't change processes easily). There’s all this buzzword bingo about “industry 4.0” and smart connected factories, but very little light on how you actually connect and upgrade factories.

In this episode we drop the buzzwords, discuss the current REALITY of smart factories, and also share honest, no-BS lessons from the process of transforming them:

00:30 — Most powerful use cases for visual inspections with AI, and why it's needed

13:10 — In-depth breakdown of the process of bringing AI to a factory

23:20 — Challenges in convincing conservative customers that AI is a journey

26:30 — Cultural differences between old-school engineers and the new cloud culture

31:00 — Why Sentin doesn’t believe in 100% self-serve AI tools for customers

33:00 — Current state of IT systems in most factories, from Christian’s perspective

40:00 — Hard-earned lessons from selling AI transformation

46:00 — Navigating the complex competitive landscape for AI and computer vision in manufacturing

56:00 — Christian’s view of predictive maintenance solutions and their claims

(Full episode with VIDEO on: https://aman-agarwal.com/age-of-ai)

(Ethics Policy: These opinions are 100% my own as an independent observer and educator. I don't own stock in guests' companies or their competitors, nor do I get paid by them in any form for any reason at the time of publishing, unless specifically stated. Episodes are also not intended to be an automatic endorsement of any company or its products and services.)

For most people, "health" is not something you can measure, and nutrition and medical therapy are two different things — you try to "eat healthy" as much as you can, and get treatment from a doctor when you're "sick."

But there's also a middle ground where you can cure or reverse certain issues just by changing your lifestyle, such as food, sleep and exercise.

So how does AI come into this picture?

My guest is Dominik Burziwoda, CEO of Perfood — which uses AI to continuously monitor blood glucose levels and a host of inflammation markers, thus being able to give patients precise alerts and specific recommendations on which behaviours are most healthy to them.

Interestingly, they sell this technology primarily to insurance companies instead of direct to consumer!

We sat down to break down the science, technology, and business of glucose monitoring, as well as other challenges and noteworthy developments in the healthcare space.

If you're serious about health, fitness or technology, be sure not to miss this episode! (Full post on: aman-agarwal.com)

00:19 — A new paradigm of “digital therapeutics,” using AI for nutrition vs prescription

06:30 — Tracking inflammation and other biomarkers in your blood (correlation with continuous glucose monitoring, time series)

09:20 — How Dominik reversed his path towards type-II diabetes without drugs

13:00 — How Perfood's AI system recommends which foods to eat and which behaviours to change

19:00 — The challenge of "describing glucose curves" and why it's Perfood's core IP.

24:30 — The evolution of glycemic monitoring technology

38:00 — Regulatory expenses for digital therapeutics, and making them "clinical grade"

43:00 — Why breast implants transformed EU regulations

51:00 — Perfood's unique distribution strategy and 3-fold marketing (insurance companies, physicians, and patients)

59:30 — Innovation cycles in pharma vs digital therapy, and the changing landscape of European VCs

1:04:00 — Near-future outlook on the medical industry

There’s an overwhelming explosion of AI products and APIs these days for so many audio-visual content tasks — be it voice recognition,  transcription, translation, facial detection, etc etc.

If you’re a company trying to use as much AI in your company as possible, you could either build these yourself, or buy ALL these tools separately, OR…. the most convenient would be to just use one API/tool that combines them all under the hood, which is what we’re going to talk about today!

My guest was Eugen Gross, CEO of aiconix. They act as a single platform for as many state-of-the-art ML models and APIs (offered by all the top providers from Google to  Baidu) so that you only have to deal with one company for all your ML needs.

What I really like about their product is that it can intelligently combine the results from different providers (say for real-time translation)  and give you the most accurate result. They’ve even been used to transcribe Angela Merkel’s speech on live TV in Germany.

But building such a company means you’re almost offering hundreds of products and services from the start, so I was curious how Eugen’s team went about their business development journey.

As always, I bring you the most thorough and insightful discussion on the topic than anywhere on the internet:

00:30 — An easy tool for AI-enhanced digital content
06:30 — How Aiconix encapsulates both 3rd party and proprietary ML models
15:00 — Aiconix’s business structure: billing and services
19:30 — Sales strategy learnings: Work directly with end customers or sell to software developers first?
24:57 — Turning away customers and refusing to build certain things
27:51 — Journey of the company and accidental pivot during Covid
32:45 – Aiconix’s unique fundraising journey with both public and private money
41:21 – Standing out in a crowded market of AI tools
43:15 – Being a sales-driven or product-driven company?

Learn more on https://aman-agarwal.com/

As a company, the number of invoices and receipts you have to deal with often grows exponentially with your business activity.

While that's good news on one end, it also means a lot of grunt work and precious human resources being spent on data entry. Which means it's a beautiful use case for automation. And while we might think, "oh just use an OCR for that," it turns out that invoices are so diverse in their formatting, that there's a joke in the accounting community that "reading an invoice is harder than driving a car."

Solving exactly this problem is Paperflow, the company of my new guest, Carsten Nørrevang (thankfully you can learn the right pronunciation of his Danish surname in the episode!) — and I learned some very interesting things through the discussion.

Here's their value proposition in a nutshell: while most OCRs and data capture software can do generic tasks, training a specialized algorithm to read millions of types of invoices is extremely expensive. Carsten said they not only have a team of 30+ people solely dedicated to data preparation but have also built special tools to help mitigate this problem.

What I find really cool is that Paperflow is a really great example of how we can use AI to fix inefficiencies across the entire value chain of today's businesses. Here's what else you'll find in this episode!

00:30 — What Paperflow does, and Carsten’s background as a former CFO
09:00 — Finance teams spend a surprising amount of time on document processing
11:30 — Why Paperflow is named, well… “paper flow”
14:00 — Solving the tough technical challenge of processing invoices
25:00 — The future roadmap: a smart document assistant
30:00 — Competitive landscape in document processing and data capture, and Paperflow’s approach
35:30 — Why banks in the Nordics are under pressure
40:00 — Paperflow’s fundraising journey
44:30 — Carsten’s perspective on going public

Full Episode here: https://aman-agarwal.com/2021/12/22/b2b-search-engine-valuer-ai/

For most CEOs, the way they hear about new technological developments in their industry is through happenstance — we read something in an article or on social media and then dig deeper, to see if it's something we should know about.

Often, by the time we find out about it, it's already "old" news and everyone's chatting about it.

Turns out, this is a fairly very passive and reactionary way of staying up to date with innovation in the industry. But there's also a way to do this actively, which brings me to my guest today!

Dennis Poulsen is the CEO of Valuer AI, which is a B2B discovery engine — it uses AI to scour the internet and see what's going on in various industries, looking for "signals" of new technological innovation.

It can then tell you in advance, things like "here are some companies similar to you, which are experimenting with these new technologies." Almost like how Spotify suggests you new songs based on music you've liked before.

I think that's pretty cool, but I can also see that it's super hard to implement, so I sat down with Dennis to dig into how their platform works! Another interesting fact is that Dennis took his company public a short time ago, so we also discuss his thought process around that decision and other things!

01:00 - Valuer's "B2B search engine," and who uses it
11:10 - Using ML to build a tree of B2B knowledge
14:25 - How the software finds "similar" companies
17:45 - Company's customers utilizing the platform to refine the searches and semantics
19:37 - Valuer as the Spotify of the business world
21:15 - The challenges in data gathering
24:23 - Data Team as key employees of the company
26:44 - Job boards predict where the company is going
27:35 - How they started and their journey as a start-up company
31:41 - Valuer's sales and marketing process
39:40 - How Valuer pulled off their "enterprise first" sales strategy, and lessons learned
47:20 - Why Valuer went public, lessons learned, and future funding plans

The manufacturing industry, while being extremely diverse (from cheese to smartphones to sportswear), is also quite uniform from a business perspective:

You're just trying to transform raw materials into a certain product, as cheaply and quickly as possible.

There's a never-ending race to do things better and more sustainably, by eliminating waste. Naturally, there's all this buzzword bingo about "industry 4.0" and smart connected factories, but very little light on how you actually connect and upgrade factories.

In today's episode, I sit with Bertil Helseth, the CEO of Intelecy, a platform that allows factories to use AI to identify opportunities for improving their outputs, and eliminating "waste." We discuss the ground realities of the manufacturing industry, and how AI is helping!

01:30 - Intro
02:36 - Using ML to fix manufacturing inefficiencies: cheese production case study
13:52 - Helping manufacturing engineers build ML models with the click of a button
19:37 - The importance of defining “normal production” in manufacturing
20:06 - The types of data available in a typical factory
26:09 - The pace of industry transformation: choosing the right factories to work with
37:30 - Challenges of onboarding a factory to an ML platform
44:21 - Intelecy's journey and how Bertil started the company
48:42 - Cement as the most electricity extensive industry
49:48 - Intelecy’s fundraising strategy, differences between US vs Europe
59:39 - Bertil's learning experience and wins in starting up Intelecy

"Electronics circuit design."

Most engineers, even some hardcore techies, don't get a feeling of bubbly enthusiasm when they hear the aforementioned words. (It's also an awkward conversation-pauser at parties if asked what you do — just say "I make electronic gadgets," jeezus!)

Because it's not fun for most people, including engineers (and because to say that the electronics business is "huge" is laughably modest), therein lies an opportunity.

In this episode, I speak with Tobias Pohl, the CEO and Co-founder of Celus, which builds AI-assisted tools for circuit design.

The goal of their tools isn't to replace the engineer, but to reduce the time they spend on certain design problems from weeks to hours.

We dive into the structure of the industry, how the electronics engineering process works (in simple English), and what it really takes to build USEFUL AI software for the purpose:

1:49 — Why engineers hate electronics!
03:33 — The current process of Electronics Engineering, and why it's so damn hard
13:00 — Nerd ego, and why “experience” in electronics design is a double-edged sword
19:59 — What can you automate? The value of machine learning in the engineer's process
28:47 — How machine learning streamlines and upgrades circuit designs
41:16 — Finding the right source of data for training Celus' ML models
45:37 — The challenge of synthetic data for ML
49:00 — Celus' journey as an AI software startup
56:06 — Celus' target market and sales process: do big companies like Apple and Sony have their own electronics design tools?
01:01:40 — Why the "build vs buy" dilemma is not a problem for software startups anymore

Today we cover a very futuristic, sci-fi-sounding application of AI!  Imagine if, just from hearing your voice, an AI application could diagnose many diseases and accurately understand your emotional states.  

Sounds fancy, and I was as skeptical as you are, but turns out it's possible. Our voice packs a surprising number of biomarkers, which can be tracked to detect issues that are difficult or impossible with other diagnostic methods!  

My guest was Dagmar Schuller, the CEO of audEERING (yes, it's spelled that way). 

 If you're as curious as I am, here's what we cover:  

03:00 — Background of audEERING and overview of their core products

 16:06 — audEERING's go-t0-market strategy and reasoning

19:07 — How does their technology work: technical overview 

23:48 — What kind of diseases can be diagnosed using voice AI? 

30:14 — Using AI to teach autistic children to convey emotions better 

32:02 — Why European medical tech tends to be safer 

36:56 — The competitive landscape for medical tech startups in Europe vs USA/Israel, and how that affects strategy 

42:12 — audEERING's interesting fundraising story: starting with only €7500 in the bank! 

50:05 — Dagmar's worst and best parts about building audEERING so far!

Not all parts of running a company are considered "sexy," even if they're absolutely critical. One of these is project management.

PM finally seems to be getting its AI upgrade, and I brought on Dennis Kayser, the CEO and Founder of Forecast, which uses ML for project management and work automation. Their system can automatically plan out a project, put it in the right  timeline, assign the right people and do many other things using AI.

We discuss:

• 01:36 — Skepticism about the need for innovation in PM software
• 8:02 — Good vs bad PM, using data vs gut feeling
• 10:25 — Forecast's customer types
• 14:14 — Why AI is transforming PM
• 20:09 — How Forecast's AI system is trained
• 27:21 — How Forecast's AI deals with the huge variety of data
• 31:19 — Solving the tricky problem of "predicting" how long a task will take
• 35:32 — How to smartly track people's progress on tasks
• 37:20 — How Forecast evolved as a B2B AI company
• 42:11 — Why a lot of non-profits collect enterprise PM data
• 43:15 — Investor outlook
• 45:21 — The right organization structure for heavily ML-oriented products

In today's episode, we'll introduce to you the new generation of voice-controlled apps and devices!

We've all seen voice assistants explode into the consumer space recently (Siri, Alexa, Google Assistant etc come to mind). But their adoption has been slow, because they feel very "unnatural" to use — it feels like a walkie-talkie conversation where 2 people take turns to speak.

Besides the inherent delay in turn-based dialogue, there's also a lack of continuous feedback to the speaker (reassuring them that they're being understood as they speak). Eliminating this discomfort is the holy grail of voice control technology.

My guest is Otto Söderlund, the CEO of Speechly — a company trying to fix that discomfort using their patented technology.

Today we're talking about machine learning on the "edge" (i.e where devices collect their own data and learn from it themselves, without having to send data to the cloud or back-end for processing).

This is The Age of AI Series, where we talk to the foremost entrepreneurs and innovators around the planet using ML to transform industries.

My guest is Jon Lindén, CEO and Co-Founder of Ekkono AI, a software company at the very forefront of this space. They provide the software and tools needed to implement machine learning at the edge for connected devices, and they're already serving customers in a variety of industries.

In recent years, there's been a lot of buzzword-bingo about the  "Internet of Things" (IoT), with people making outlandish projections that everything from your toaster to your teacup will be connected to the internet by 2025.

Jon and I cut the buzzwords and get down to the actual facts —  explaining the actual IoT and edge-learning use cases we're seeing in  the market, what it really takes to build good applications, and how  quickly or slowly the landscape is changing:

• 01:13 — Concrete examples of Ekkono's customers and what they're building
• 03:23 — "Predictive maintenance": quick case study on heat exchangers and fluid handling
• 09:41 — Where is the value in edge ML for industrial applications
• 12:53 — Using edge ML for the consumer market: example of intelligent lawn mower
• 16:28 — Purchasing and investment decisions around IoT: how the perception of "value" is changing
• 18:02 — The technology of edge ML explained in simple words
• 20:51 — What Ekkono provides to their customers and what they use it for
• 25:22 — The building blocks of a modern IoT application, and how it works
• 31:40 — Why the IoT space hasn't matured quickly enough: an unbiased perspective
• 38:00 — "Every billable hour is a failure": helping buyers get started on their IoT journey, without having to become a services company, and dealing with competition
• 43:16 — Investor sentiment towards IoT companies
• 47:26 — Aman's summary of the episode
• 49:48 — Jon's comments on Aman's summary

Here's the full episode link: https://sanpram.com/2021/10/17/edge-ml-ekkono/

Traditional news media companies have struggled to shift online — ad revenue is dropping, and they're even competing with millions of independent blogs and social media feeds for readership!

This is The Age of AI Series, where we talk to the foremost entrepreneurs and innovators around the planet using ML to transform industries.

My guest today is Mari Ahlquist, whose company iMatrics provides AI software that intelligently tags and categorizes news articles, to enable personalized recommendations to readers like Netflix or Youtube.

• Setting up the problem:
  01:44 — The real challenges of running a news media organization in 2021: revenue, subscriptions, etc.
  07:12 — The "Coca-Cola problem" that news media faces in terms of competition
  10:34 — Why structuring, categorizing and tagging content is the key technical feature that enables all other things
  
• Why tagging is a hard ML problem, and approaches to solve it:
  17:38 — Why tagging is an ML problem, and can't just be programmed
  20:38 — Potential landmines: how your tagging system can affect the creative freedom of journalists, and how to avoid that from happening
  27:40 — Exactly how "intelligent" does the algorithm need to be to tag articles better than humans?
  29:34 — How the product really works (simple explanation), the huge number of corner cases, and why the system has to be surprisingly complex
  39:19 — The world is changing. How do the automatic tags keep up?
  42:05 — Aman goes into a geeky brainstorm about predicting the coronavirus pandemic and Mari politely indulges him
  
• Getting operational: how the team and product evolved
  47:15 — How many engineers they needed to build the system, how much is manual vs automated, etc
  50:00 — Origins and product evolution
  

Here's the original blog post: https://sanpram.com/2021/10/09/ai-newspapers-netflix/

This podcast is produced by Sanpram Transnational Corp., an education and media company helping non-technical professionals be tech-fluent.  See our programs: https://sanpram.com/teaching/