Looking for a really smart AI assistant? You’ll be surprised.

Photo by Alex Knight.

This morning I’ve read two very similar and interesting stories. Their authors told about one of the today’s most overhyped topic — personal AI assistants. The both stories are critical and skeptical. They doubt the ability of these “smart” bots to augment human intelligence in an effective manner.

It might sound a bit strange but I agree with authors. As a founder of AI-inspired startup Summa — I’m sure about the big potential of AI in assisting humans. Still, it would be too optimistic if we oversee all limitations and risks of using AI assistants.

Real people behind your “smart” AI assistant

The first story “The Humans Hiding Behind the Chatbots”, delivered in an amazing daily email by Sam DeBrule, was published several days ago. As a practitioner, I certainly know that every AI “brain” need manual human setup and training. Thus the fact that “humans had to be on duty at all times” doing a big part of AI’s job behind one of these “smart” bots was very surprising.

Illustrator: James Singleton/Bloomberg

The story is full of other interesting facts and thoughts. It raises a very important question about our privacy protection by using any AI assistant today. It might look very attractive and cheap, but in fact companies “use similar and vague marketing language and don’t often divulge operational details”. Even if trainers should deal with anonymized data, who knows if they don’t want (and can) de-anonymize them in any way?

Or, as another former X.ai trainer put it, he wasn’t worried about his job being replaced by a bot. It was so boring he was actually looking forward to not having to do it anymore.

Today’s AI assistants have a long way of progressive evolution to become really smart. Ideally, they must be quickly trained and become autonomous— so we don’t need constantly supervise them as small children.

Unexpected consequences of using AI meeting schedulers

The second story by Alex Fishman is written more from the practical point of view. His bottom line isn’t inspiring for AI assistants creators.

AI assistants are too dumb to trust right now and they’re too dangerous to ignore.
Janko Ferlic

To be honest I am not so extremely skeptical about AI-bot scheduling the meetings. It can be very valuable for some kind of professionals (like HR pros etc) who deal with lots of meetings day by day. It doesn’t look so strange to receive bot’s message when you are seeking a job. You have to follow the existing hiring rules — and to discuss the meeting time with a bot.

But this use case may not look very farsighted for other cases, especially for networking and everyday communication. Humans like to converse with humans. Today, no bot is likely to withstand a meaningful conversation with human more than several seconds. You do not want to chat with them after few dumb misunderstandings.

Don’t get me wrong, I understand the need to improve productivity. However, when forcing a tool on your business partners and customers, think about how it may affect them and consider the message you’re sending by doing so.

P.S. My founder’s takeaway from these stories is pretty simple. Putting AI bots in the front of any human-to-human relationships is a very risky idea. No AI bot today is capable to take into account all nuanced contexts and meanings involved in real world communication. They may assist or augment us in the desired manner, but they cannot replace. Humans are not replaceable.