I've built a prototype AI model that helps buyers make a better decision on what boat

[DingiesnTheories]

When this model can encapsulate every vessel, make, model, etc...

Would you consider using it to help you buy your next boat?
Do you think your fellow boat lovers would use something like this for their next vessel purchase?

Any other thoughts are appreciated - just a little background on why I built this - my friends and I we're thinking of going in together on our first boat but we had no idea what we wanted to get. The prototype works quite well and I'm just looking for some market validation before I continue building this.

P.S. We ended up buying a SeaStella 55ft Flybridge

 

[READY]

Very curious how you built a model if you have not owned a boat? I have had North of a dozen and not confident that any logic would lead to the boat I purchased but all but 1 were the perfect boat (at that moment). What are the inputs to your model.

You could back test it on current boat owners. Happy to be a test case for that.

 

[Comodave]

Personally, no I wouldn’t use it. Can’t say exactly why not but I would be cautious about whatever personal preferences the programmers put into the software.

 

[huruta]

I’d be interested in the parameters of the training dataset. As someone whose work involves prediction, I’m skeptical that all the relevant parameters would be included. There’s an I’ll defined emotional component that would need to be quantified. Expert guidance in model development can be important- do you have this perspective? Data alone can’t do a great job imho.

 

[DingiesnTheories]

Very curious how you built a model if you have not owned a boat? I have had North of a dozen and not confident that any logic would lead to the boat I purchased but all but 1 were the perfect boat (at that moment). What are the inputs to your model.

You could back test it on current boat owners. Happy to be a test case for that.

The majority of the weight in the buyers inputs is based on how they'd like to use the boat - but the rest is on the backend - it's a little like Tinder or Bumble for buying boats - the algorithm gets stronger with a better buyer profile, sure, but even more so based on the decisions the user makes.

And sure, that would be appreciated - I'll PM you.

 

[DingiesnTheories]

Very curious how you built a model if you have not owned a boat? I have had North of a dozen and not confident that any logic would lead to the boat I purchased but all but 1 were the perfect boat (at that moment). What are the inputs to your model.

You could back test it on current boat owners. Happy to be a test case for that.

I’d be interested in the parameters of the training dataset. As someone whose work involves prediction, I’m skeptical that all the relevant parameters would be included. There’s an I’ll defined emotional component that would need to be quantified. Expert guidance in model development can be important- do you have this perspective? Data alone can’t do a great job imho.

See the other reply but if you're interested in learning more, PM me, I'd love to talk more.

 

[huruta]

Hard to quantify parameters come to mind: value on locally manufactured boat, boat reputation, ‘hold type’.
(We don’t like to buy big ticket items often….don’t want to buy up frequently but rather buy for longer term to start…how to quantify ?). Demographic characteristics of owners of current brands could be very useful, perhaps you have this already? Go big with a great, strong predictive model or don’t bother. Lots of lackluster tools undermine confidence in outputs.

 

[Comodave]

The first thing I look at in a prospective boat is how it looks. The lines of the boat have to be pleasing to my eye. How will you quantify that in a program? That is one of the reasons why I don’t think it would work, at least for me. Life is too short to own an ugly boat…

 

[SteveK]

Personally, no I wouldn’t use it. Can’t say exactly why not but I would be cautious about whatever personal preferences the programmers put into the software.

I have never regretted any boat I bought, so how would I know if this app works. All my boats fulfilled a want more than a need.
There are plenty of newb posters here asking for help find me the right boat, target audience perhaps.

 

[Comodave]

I have never regretted any boat I bought, so how would I know if this app works. All my boats fulfilled a want more than a need.
There are plenty of newb posters here asking for help find me the right boat, target audience perhaps.

Me too. We are on number 24 and some of them if I had just gone by the specs and photos I would never had purchased. But when I got on board they just felt right. No computer program can do that.

 

[Pete Meisinger]

I love boats! I love owning them and I love buying them. Alas, I also love selling them, but that is a different story.

In buying a boat, I would think most, if not all buyers have "Boxes" which must be checked. Some might formally use a spread sheet, others might use a yellow lined pad. No matter, certain criteria must be met. If the buyer wants a "Up galley" a "Galley down" will not work. Same for flybridge vs pilot house, same for single vs dual engine, same again for sail vs power. The list goes on and on, literally forever.

If a buyer has done his diligence the choice becomes fairly clear, fairly quickly. They have, in effect, built their own program.

Do they need yours, probably not. Would anyone use it? Maybe for fun, if it were free.

pete

 

[MYTraveler]

Is it truly AI? If so, what data set did you train it on? The concept is so inherently subjective that I can't help skeptical about the availability of a large enough data set to make AI reliable in this context. As such, to answer your question, my use would be limited to identifying boats and features that I may have failed to consider.

 

[twistedtree]

I'd be interested to try it and see what it recommends. If nothing else, it would be an interesting test and probably useful feedback for you. There are probably others here who would be willing to give it a try too.


That said, it's not something that I'd ever be inclined to turn to for guidance. All my experience with AI reminds me of 1990's support calls with outsourced call centers. The person seldom actually understands the question, but answers anyway with information that is typically incorrect or not helpful. All it does is waste more of your time. Part of the problem, not part of the solution. That's not something I voluntarily seek out, in fact, just the opposite.

 

[DDW]

How emotional does your AI engine get? Because buying a boat is an emotional decision, not a rational one.

 

[tiltrider1]

I see an interesting tool. I see a very difficult challenge. I am curious to see how the tool performs.

How does the tool know about every make and model boat out there? I doubt I would let the tool make a decision for me but I would certainly consider its opinion.

 

[Northern Spy]

I'm with Commodave and DDW.

There must be a "look over your shoulder as you leave your boat" component to the algorithm.

 

[SteveK]

I'm with Commodave and DDW.

There must be a "look over your shoulder as you leave your boat" component to the algorithm.

You don't have a sig line, I think you just made a good one.

Every time we leave the boat it gets a last look back.

 

[Comodave]

You don't have a sig line, I think you just made a good one.

Every time we leave the boat it gets a last look back.

And it has to look good to your eye…

 

[FWT]

I think you are asking the wrong audience

The Facebook pages where people who have never boated ask: “what boat should I buy to do the Loop?” would be the better audience.

 

[psneeld]

Something like 31,000 members and really only a handful of active posters.

Bet there are a few out there that aren't so salty that they can rattle off a buying list with pages of why and why not for each boat with serious experience to back those reasons.

But true, there are many other venues full of dreamers that need help.

Wonder if this AI has the answer to all the questions here that get debated endlessly....best size to singlehand, single or twins, galley up or down, liveaboard and sell dirt or not...?

 

[DDW]

Perhaps more useful would the an AI based program that recommended the best anchor....

 

[Barrie]

All my next boats were based on things i liked and things i didn't like on the previous boat. Experience. One boat lasted 2 months. Others years. It is not until you get "in bed with the lady" that you know you done good or not.
Barrie
The six stages of life: Toyboat, Sailboat, Motorboat, Motor Home, Nursing Home, Funeral Home.

 

[Duetto]

just curious what's it written in? does it use an off tghe shelf inference engine?

 

[La Sirena]

Behind every program is a human. I don't want to put my hard earned money on another human's opinion, especially when I haven't met them.

 

[RT Firefly]

Greetings,
Mr. LS. No worries. When AI becomes sentient we'll all be driving these vessels.

 

[Art]

All my next boats were based on things i liked and things i didn't like on the previous boat. Experience. One boat lasted 2 months. Others years. It is not until you get "in bed with the lady" that you know you done good or not.
Barrie
The six stages of life: Toyboat, Sailboat, Motorboat, Motor Home, Nursing Home, Funeral Home.

Then Final Home... The Big H - AC'd with fine food, fun learning events, good souls sharing dialogue... or... Hot as HELL with DT and followers forever continuing to spout off BS rhetoric!!

 

[MYTraveler]

Well, OP hasn't been back. I guess we have outlived our usefulness, but I was hoping he would answer my question about the dataset on which he trains his AI. His idea of using AI software to help select the perfect boat isn't viable, IMO, primarily because the validation data just isn't available. If the concept worked, there would be a much broader market for things like helping to select the ideal career. Of course, there is the Kuder Preference Survey for that, but it doesn't use AI and instead seeks to correlate a subject responses with those who report being either exceptionally satisfied, or dissatisfied in their careers. I can't imagine how a good AI data set would be collected for something like that, and was really curious to hear OP's answers.

Of course, these days it is fashionable to slap the AI label on everything, regardless of whether the underlying technology makes any actual use of AI, just as renaming XYZ Company to XYZ.com was in vogue 30 (?) years ago to hugely increase a company's market cap without any actual change to the business plan, beyond a vague intention to create an internet presence accessible through a website at XYZ.com.

 

[FWT]

Just a thought.

It might be more realistic to build a searchable database of all known boat models. Searchable by age, size, draft, air draft, beam, pilothouse or sedan, flybridge or not, engine characteristics (twins or single, fuel, brand, etc), berths. Include a few pics. Build out the field of view, then let users narrow things down to what looks interesting.

And, if you can't build out the database, then you can't lead anyone to any useful direction. So its a necessary first step anyway.

A lot of work, to both build and maintain. Hard to see a payoff for the effort.

 

[SteveK]

My previous boat the Grand Banks classic 36, considered the original trawler, often copied may be interesting for this AI to come up with the right one.

In my search I wanted a 35-40 and so looked at what was available within my chosen price range. The GB popped up, a long time desired boat. So I looked at dozens of boats as a comparison to the one I knew I was getting. None changed my mind, several helped decide.

The tri cabin, aft stateroom, two heads, dual station, 36 footer boats seem endless.
Under what criteria then would this AI choose the boat to buy.

I find this similar to an aptitude test to determine if you are likely to be successful in a career. I was asked to take one, I proceeded to answer the questions that pointed me to be perfect for the job, got the job and did well.

 

[MYTraveler]

I find this similar to an aptitude test to determine if you are likely to be successful in a career. I was asked to take one, I proceeded to answer the questions that pointed me to be perfect for the job, got the job and did well.

Sounds more like you took a preference test, than an aptitude test. Probably correlated, but designed to assess two different things.