Goodbye, AI ?? ?? The hype is over.

Just kidding.?The AI hype is alive and well, at least judging by how much VC funding is going into AI startups - a whopping 28% of the total, 8% increase from the previous quarter - all while VC funding volume dropped by?7%?Q over Q!?

So since AI Products are getting more and more important - becoming an expert in AI Product Management and Go-To-Market will be too - because as we know, AI products still have many, many issues - both technological, product-related, and...ethical (I'm not kidding. Don't roll your eyes. Keep reading and let me know what you think).?

Luckily, we have two amazing speakers talking about AI Product Management and Go to Market on our next Product Drive in October (it's a FREE online conference, for the new folks in the room) - join 3,650 folks who already signed up!?

So what challenges does AI pose for PMs?

Watch my video TL;DR?

The Tech Challenges?

Well, the first problem with AI startups is that they are like a bad boyfriend - they promise you the world, and then - rarely actually deliver on those promises.?A year ago, one of our competitors (the big, expensive one) make a big, bold, loud product announcement of their AI-powered product analytics and in-app engagement experience builder.

I was anxious. It sounded too good to be true -automated insights so you don't actually need to have any training in data analysis to understand your user's product usage data. And then - in-app guides that just build themselves - based on those insights.?

Fast forward a year, and there's no sign this revolutionary product is actually going to materialize. So what the hell happened???

Large Language Models don't actually think - and training them to perform such complex actions generating user onboarding walkthroughs without actually understanding the use case behind your product, as of 2024, verges on the impossible. And as a PM in an AI startup - you may be sometimes still asked to do the impossible...

For creating walkthroughs that are possible to create - check out Userpilot !

Stability of the models and quality of the results is a massive engineering problem that most AI companies haven't been able to overcome. We've all seen "funny" generative search results - but they are only funny for us, tech people, as memes, not when you're a vulnerable person looking for medical or financial advice...

There is a host of problems to overcome:?

• Scalability: AI models need to scale effectively within the product's architecture. PMs must ensure that the AI components can handle large volumes of data and interactions without degrading performance.
• Data Collection: AI models require vast amounts of high-quality data for training. PMs must ensure that the data is relevant, diverse, and representative of the real-world scenarios the AI will encounter.
• Data Quality: Poor-quality data can lead to biased or inaccurate models. PMs must work closely with data scientists to ensure that data is clean, annotated correctly, and free from biases. "In my opinion, the biggest challenge is data confidence.?If the AI provides wrong answers, the user will lose trust in it and never return to the product." - said Sonia Piorek, ?associate Product Manager at VC lab.?And what's worse, according to our Head of AI - the models' performance can actually get?worse, rather than better - over time:? ?

And the big question, which Bart Jaworski -?Senior Product Manager at Stepstone & Product Management Coach?- asked in our private conversation recently - is - are these problems even worth overcoming for your market, or are you just following the hype and building features nobody needs??

The Usability / Product Challenges?

A lot of the VC funding going into AI startups today will be spent on useless features nobody wants or needs. A lot of it will be scrapped. Aside from the ethical aspects of wasting all that money...what happened to good, old product discovery??Like - I'm sure nobody asked for an AI-powered chatbot with a widget that listens to all your conversations, and then transcribes them and sends you creepy messages to keep you company...

It's useless. It's dangerous (data security anyone?). It's creepy as hell.?And yet - according to crunchbase -it just raised a $2.5 million seed round.?

How much good could you do with 2.5 million dollars??This brings us to my final, but perhaps most important point...

Ethical Challenges

I've ranted about ethical challenges in Product Management before - building addictive or useless or downright harmful products and "dark patterns" have been around forever - but "AI" have taken these challenges to a completely different level:?

Data Accuracy:

Since we know that LLMs are by default prone to error, how do we deal with serious implications of misleading or inaccurate information they give? Inaccuracies can lead to harmful decisions, biased outcomes, and erosion of trust, particularly when AI is used in sensitive areas like healthcare, law enforcement, and finance. Who will be accountable for potential...manslaughter caused by AI errors??

Bias and Fairness:

AI models can perpetuate existing biases in training data, leading to unfair outcomes, particularly in sensitive areas like hiring, lending, or law enforcement. The fact that the ML models are mostly trained on English data - is also adding to the cultural discrimination, exclusion, and even potentially, over time- cultural extermination...Read this brilliant post? by?Artur Kiulian ?on the impact of AI usage on non-Anglo-centric cultures.??

• Privacy Concerns: AI often requires vast amounts of personal data, raising concerns about how this data is collected, stored, and used, potentially leading to privacy violations. Now imagine some AI gizmos are listening to all your conversations...and someone hacks them? ?
• Ethical Use: The potential misuse of AI in areas like surveillance, autonomous weapons, or deepfakes can lead to significant negative social impact. I keep reporting all the deepfakes on social media, but I know the fight has been lost. We'll soon be flooded by perfectly unattainable images of deepfake influencers, further distorting body images and ruining the self-esteem of millions of young girls (and boys too.) See what I mean??

Don't get me wrong, I'm not a luddite. I'm using chatGPT literally all the time (probably shouldn't for the sake of my math skills ??) The AI investment hype should not be over - it's a good job we are investing in such a life-changing tech. But I believe we should be dedicating the same resources to doing proper discovery work ("Is this AI solution really needed, or are we building it just for the AI's sake?"), data security and addressing the ethical considerations.?

What are your thoughts on these Product Management challenges or AI product?

Join us for Product Drive and let's have a LIVE debate on this!?