ZF Lee
Legendary Contributor
FASTLANE INSIDER
EPIC CONTRIBUTOR
Read Rat-Race Escape!
Read Fastlane!
Read Unscripted!
Speedway Pass
Today in my forecasting lecture, my professor said,iam very interested in this topic, here are my questions for you AI pros out there:
1. What business in AI would be profitable if i start today in 1-3 yrs?
2. I guess coding is a big part, so i cant do it alone, how would you solve that problem if i want to start a AI business with low budget.
3. What fields are most affected in the next 5 years by AI! please give me some answers i cant google haha
Cheers
'I don't think AI will totally replace the jobs. But what is certain is that it can automate certain components of the job.'
He explained that because at that moment, we were looking at R programming codes to use for ARIMA, an approach that needs us to look for white noise, that tells us whether the data is legit or not.
We can automate the creation of the graphs, but we still had to read it and its benchmark indicators to tell us a story.
So, 1-3 years is a short time.
Me thinks that to pick the lower hanging fruit, you have to see:
1. which job component can be directly automated- maybe its currently undergoing automation already, but still has room for expansion
2.which data that the AI (or stats software, for that matter) can pick up easily, or can be coded in smoother, with faster data cleaning and pre-processing.
I think you'd have to consider what most programmers of AI can simply do, on average, since they will still be doing some work.
3. Type of deliverables. If your ultimate product is info/data based (e.g. AI-based market report), then you can quickly market that solution. But if you have to use the AI results to transfer info to let's say, for a physical factory mechanism to work, that of course, it takes some costs and time for implementing.
IMO, probably accounting (especially costing and logistics) and finance would be most affected. Much of it is info/data based, stuff you can transfer quickly on the cloud as deliverables, and the observations values are mostly dollar-sign....
Fundamentals like costs, optimization, risk and volatility, are inherent in basic statistics behind AI, and anything related to the transaction of money is directly impacted.
The two fields are already affected, but somehow we are still stuck with archaic methods like banking checks...the final nails in the 'coffin' are yet to be delivered.
IMO, I think there's room for AI, or software tools, that:
1. Not only LINKS psychological and demographic info with financial data for regressions, forecasting, etc.
2. but also ASSESSES and UPDATES the models consistently, as there is never one perfect model-especially the psycho and demograph part.
On a side note...
There was an earlier argument in another thread between James Fend and JScott as to whether data methods trump human emotions in telling what comes in the future, and vice versa.
Both have good points, but I think if we can throw things up into an equation for analytics, it might be good to take the error term, and see if we can get some constant or a function to capture some of that error in terms of psycho/demographics.
Of course, that so-called function or constant would be dynamic, depending on the parameters of choice, but that's where automation can come in.
As for roping in help, you can try looking into the open-source model.
I found it interesting that the R program environment is almost entirely open-source generated, as well as its packages.
I have mentioned earlier that there is a website on this thread called Kaggle, where folks join contests to come up with the best predictive models for a problem.
And while there are cash prizes, somehow, reading their immense documentations feels as if the folks who join it are mostly there to do their passion for data, rather than for the cash alone.
Is it possible to call on such people, network with them? Maybe.
Dislike ads? Remove them and support the forum:
Subscribe to Fastlane Insiders.