Are you talking to a human or an AI?

Good morning

It is time for another newsletter filled with thrilling news, practical use cases, and scary developments in the AI industry.

It has been an intense week, so much so I needed to send you guys an extra newsletter.

I think today’s newsletter will be interesting but as always a bit concerning as well.

Let’s jump in…

I’m calling to book a woman’s haircut for a client

Imagine for a moment that you work as a hairdresser.

A customer calls to make an appointment.

She tells you what day and time she wants her hair cut.

You check the calendar and suggest a time.

The customer asks for another time.

You check again.

You find a free slot, and you agree with the customer.

The customer gives you her name.

Booking confirmed.

Call finished.

You go back to cutting hair.

Could you confidently say that the conversation you just had, was with a real human being?

Or is it possible it could have been with an AI?

It is just a matter of time before mundane tasks can be outsourced to an AI, including things that require you to converse on the phone.

In 2018 Google showcased one of its AI developments.

Tell me, honestly, is this not blowing your mind right now?

I know I said in the unscheduled “breaking news” email on Tuesday that Microsoft is winning the AI game.

But we should not discount Google.

At least not yet!

They are simply holding their cards close to their chest.

No doubt they have a bunch of “out-of-this-world” AIs hidden away in their secret AI lab!

Conversational AI and Chatbots for enterprises

Most of us have had the “pleasure” of chatting with a chatbot.

You find them on many eCommerce websites, insurance companies, banks, and more.

Many are using “older” and basic chatbots with limited functionality.

They can handle a limited number of tasks and use rule-based programming to match user queries with potential answers.

These chatbots halt as soon as the customer asks a question that has not been defined beforehand.

Enter conversational AI.

This AI understands human language and can perform more complex tasks and help customers better.

Conversational AI uses a powerful AI with deep learning capabilities and natural language understanding.

And the AI can be trained.

Conversational AI, in short, is the synthetic brainpower that makes chatbots capable of understanding, processing, and responding to human language.

Imagine the complexities of challenges a chatbot faces:

• Grammatical errors.

• Spelling mistakes.

• Wrong/non-use of punctuation.

• Limited vocabulary of user.

• Users not knowing the “terminology” and using natural language to explain their problem.

• Multiple language variations (in Norway we have two language variations).

• Slang.

• Filler words/not being concise in your formulation.

• Different languages within one country (like Belgium, which has three official languages).

I am sure there are many more, but these are on top of my head challenges the AI must be able to deal with.

With the advances in conversational AI, I think it is safe to say that AI will disrupt the customer service industry.

Let’s have a deeper look!

Applications and implications

Conversational AI is used to power advanced customer service tools.

They do come with a price tag, though.

But it is not hard to understand how a tool like this creates cost savings and positive ROI for businesses, not to mention increased customer satisfaction.

As we are entering a time where AIs are getting scarily powerful, I think it is just a matter of time before we get customer service AI tools where instead of having a chat box to enter your questions, an AI avatar pops up on the screen, and you converse with it.

Boost.ai, from my hometown of Stavanger, in Norway, is one of the market leaders in enterprise chatbot solutions.

(I am not paid to talk about them - they don’t even know about it until now…)

This picture has no relation to the story or AI, but I needed to break up the text. Plus, it was an excellent opportunity to be patriotic and show you this picture from my favorite place back home, The Pulpit Rock, (it is a 604 meters/1982 feet drop into the ocean).

In a previous newsletter, I spoke about the video AI companies Colossyan and Synthesia (https://practicalai.beehiiv.com/p/not-morgan-freeman-scary).

I can see companies like them being bought up by providers of enterprise chatbots like Boost.ai - or even the other way around, Colossyan or Synthesia buying a company like Boost.ai.

Heck, I can even imagine someone like Oracle, Salesforce, or Hubspot buying all three of them…

But I digress.

Here are some obvious use cases for Conversational AI chatbots for enterprises like banks, insurance, telco, etc.:

• Blocking a credit card

• Filing insurance claims

• Updating insurance claims

• Upgrading mobile data packages

• Upselling products and services

• Automating mortgage deferment requests

• Generating and issuing invoices

• Stopping your subscription for a month

• Asking basic financial questions to your bank

• Changing your insurance package

• Cancelling X-service

• How do I do X with product Y

• What do I need to bring to the meeting about X with the bank

• FAQ style chatbots

• Customer surveys

• Adding missing information to an insurance claim

• Asking only for the needed information in order to process X

• Booking your next appointment with the hairdresser 🙂

• Hotel bookings

• And so much more…

Right now I am in the process of filing a claim with an international insurance company.

And let me tell you, it feels like this company makes shit difficult on purpose!

I have filled out the papers twice, been on the phone with support twice, sent two emails, and uploaded two extra PDFs.

And I STILL have no effin clue if I have done it correctly.

All I have been told via an automated email is that “they have received my claim.”

So there’s that…

But the claim form is 11 pages long!!!

11 pages!

ELEVEN.

A majority of the questions on the form are not even relevant to my claim!

Or what about the fact that I am in Europe, and I want my compensation paid into my European bank account, which uses IBAN account numbers and SWIFT, not a check! (Hello, America, can you join the rest of us in 2023?).

If this big insurance company had a chatbot from Boost.ai, my claim would get filed correctly the first time.

I could have uploaded all documents and supplied all the needed data like my European bank account, bank name, etc. seamlessly.

*** Rant over ***

Customer satisfaction is another big one for conversational AI!

Happy customers stick around.

Unhappy customers leave… (yes, I’m leaving that insurance company).

AIs are efficient and will bring cost savings in staffing, offices, and more.

Just like a robot in a factory brings increased accuracy and efficiency, a chatbot using conversational AI can benefit the enterprise similarly.

So what are the implications?

As with so many things with AI, robots, and efficiencies in general, there are job losses.

But more than anything, I think these tools are a serious threat to call centers.

Imagine when voice AI gets as good as in the Google video.

Don’t you think it will be able to replace specific sales roles too?

How about those salespeople calling from non-profits to ask for your donation?

Or those people calling from survey companies to ask you all kinds of questions for consensus surveys?

I think they were on the way out already.

But with realistic human-sounding voices with an AI behind them?

Checkmate.

Imagine a company like Nielsen, Ipsos, and Kantar going all in on AI.

Come to think of it; I can see one of these companies buying up an AI company or two in the future (or getting eaten by one).

My brain is frying thinking about all the implications and uses of this kind of AI.

How about you?

Hit reply and share your thoughts!

-

I want to finish by sharing a link to the case studies page of Boost.ai.

It is interesting to read about the incredible results that conversational AI delivers at an enterprise level.

Another awesome resource you may want to check out is their Ultimate Guide to chatbots for enterprise.

Even if you are not working for or interested in anything “enterprise”, it is a great resource on chatbots, conversational AI, and AI in general.

Someone asked ChatGPT to answer a question based on IQ level

The experiment of the week - I created a podcast with AI

After last week’s newsletter on voice AI, I wanted to find out if voice AI has any practical use cases for my business.

So I bought a license for Revoicer AI, a TTS software with fairly realistic voices with moods built into them.

On their sales page, they list podcasting as one of their use cases.

I wanted to find out if I can simply copy/paste my text (my newsletter) into the prompt, and seconds later have a podcast episode ready to go.

The advertising makes it sound so easy.

But is it really?

I started Revoicer, picked a great-sounding voice (Nova), opened the Google Doc, copied the first page, pasted it into the prompt, and clicked the “Generate voice” button.

Drum roll…

It’s a fail!

In short, it is NOT as easy as copying and pasting text into the prompt and getting back a fantastic sound file ready to publish.

To get an acceptable output, you really need to work with the text.

The AI cannot or does not pronounce the “dot” in an URL, for example.

This may be a bug - or a mistake they made.

I don’t know.

But when the text says “Go to www.futurehandbook.com” the AI will read it as “Go to www *silent pause* future handbook *silent pause* com”.

This means that all website addresses need to be re-written as “Go to www dot futurehandbook dot com”.

That is just one of several issues I had.

Eventually, I managed to turn the newsletter into a sound file and uploaded it to Soundcloud.

But it took me over 4 hours to create a decent episode read by the AI.

With more training it would be faster, I’m sure.

But if I take out my fancy Shure microphone, connect it to my computer, and record myself reading the newsletter, I will have a high-quality episode with great sound ready in 10-12 minutes.

Thus, I conclude that TTS AIs like Revoicer are useless for creating podcasts.

This is not to say that Revoicer and similar tools are bad tools.

Far from it.

I think they are pretty cool.

And there are many other use cases.

But for podcasting they are a time suck and of little use - at least for now.

I am writing a full article on this experiment, and I will link to it in next week's newsletter.

Microsoft presents Vall-E - the next scary AI

A few days ago Microsoft released a paper on Github on their latest AI called Vall-E.

Let me tell you, it is absolutely bonkers crazy.

With only a 3-second clip of someone’s voice, the AI can replicate their voice.

And it is really, really good!

Like, “is this a prank” good!

I wrote an article on Vall-E over at my blog The Future Handbook if you want to read more.

I also created a 1-minute video with their test files to show you how crazy this is. Have a look and listen:

AI art of the week - Handicrafted Albert Einstein doll

Quick links - interesting AI news and stories

My friend and online marketer Gael Breton of Authority Hacker discovered that CNET, a big online publisher, uses AI-generated articles. It reached the news.

This article from Forbes argues that 2023 need not only be the year of AI but the year of AI Education. AI is a big part of our lives, and education is needed asap.

Gizmondo has made a list of over eight industries “our new overlords” plans to kill. One of them is customer service, today’s topic. Have a read to find out the seven others (it’s a slideshow, so press the arrows on the images)

DoNotPay CEO Joshua Browder posted on his Twitter that he is offering $1 million to any lawyer or person willing to argue their US Supreme Court case using their AI. The news spread like wildfire all over legal blogs like this one.

That’s all for this week folks.

Yet another long one.

Hopefully, you found it interesting and useful!

Next week my ambition is to make it a bit shorter, with more practical examples of what some intelligent and creative people have done with ChatGPT/GPT AI so far.

So unless Google releases some big news or Microsoft decides to buy up more AI companies, this is what you can expect in the next one.

If you have any questions about AI or any feedback, just hit reply or tweet me @thomassorheim

And I would really appreciate it if you would spread the word about this newsletter.

I know the link I gave in the Tuesday newsletter was broken. But I appreciated the effort (I could see from the statistics that many of you did click that link).

All issues, previous and this one, can be found by following this link.

PS! Why did ChatGPT cross the road?

.

.

.

To get to the other side of the conversation!

This is not the end. It is where the fun begins.