AI, AI, AI… Share volume of mentioning of AI scares me – not because I believe AI is currently capable of the “Judgment Day”[1] – but because I believe we do have too many talks, and too high expectation of AI, and especially Generative AI. It’s just way too much!
I wander, why do we really want AI? Do we really lack “Intelligence” so we need the “Artificial” one? Or we just admire AI as a concept… Or is it just trendy… Or something else…
Sometimes ago, I heard – there is no such a thing as “Artificial Intelligence” – either there is intelligence or there is no intelligence…
People are getting an “artificial” impression, we cannot do anything anymore without “Artificial Intelligence”… I am seeing people are under enormous stress trying to find some “really” good use of AI (Generative AI – per say). AI is so trendy, so we just must have it in every “cool” company… But lets ask the army of the people training Generative AI models? Do they think AI is so “cool” and so much intelligent?
AI has potential to be good, has potential to be bad (“Judgment Day” or so) – but it is still far away from the either – well this is at least how I see it…
What is AI, and how to make use of it?
In the essence, Artificial Intelligence represent machines that mimic human intelligence and human cognitive functions. AI is used for solving complex problems, usually and historically done by humans – the most commonly used for visual recognition, speech recognition, but also content generation[2]. But current AI still does require human action in feeding and training the model[3] – and this is exactly where current Generative AI, like ChatGPT, “fails” (my impression) – often, too much human effort is needed to train the model to imitate human cognitive capabilities… And still it does not provide full-proof results[4][5]. Will this improve over time? Possibly…
While I do question common perception of AI being “real” human style intelligence, I do not question AI as a concept. It does not mean AI, including Generative AI, should not have its place – if we remove usual “AI marketing”, there are many practical uses-cases where AI tools can help (but really help) people – as long as we are aware and we embrace current (and future?) limitations of those tools (we call “AI”)[6][7].
Let me go through few examples…
SAP Business AI concept does bring refreshing approach – SAP does not “really” try to built its own Generative AI or so – but lets you “integrate” various AI engines (as per your preference?) into BTP platform and using it in optimizing daily business process, or even development – interesting and pragmatic approach, if you ask me,
SAP Build Code[8] is another interesting concept promoted with SAP[9]. While AI in general terms tries to mimic human thinking, in the coding we humans are actually trying to “speak” machine understandable programing languages. So, could machine better “mimic” programming languages, based on human commands in the “natural” language? After all, programming languages do have definite number of possible “phrases” & “clauses” to be learned… Still coding the program is one thing – resolving complex IT problems is something completely different.
Microsoft 365 Copilot can be useful as well – e.g. can help us manage Teams Meetings and tasks more efficiently, process or build Word documents more efficiently, provide some useful recommendations… But be careful, it’s not fool-proof – I know from the first hand – so some human overseeing the work is still needed. Nevertheless, it can be really very handy in speeding-up time consuming tasks.
Also, Windows Copilot can be useful – e.g. providing equivalent to GPT-4 and DALL.E 3 (aka Bing Image Creator) capability in Windows 11… But be careful here as well, it learns on any data you submit in the queries – so forget about privacy. Is this a problem? Not necessarily – as long as we are aware of this “limitation” and do not share any private and confidential data – we are more-or-less okay.
But there are other useful tools & products, like IBM Watson, which was capable of creating great (and useful) Chatbots, much before ChatGPT – I know, I was doing that.
And that’s just few…
However, the downside of any of those “smart” AI tools are not so much in tools itself – but in the overwhelming dependency, we ourselves create, on AI tools – we already tend to trust AI tools too much – and this could take us to the parallel reality, where AI generated “hallucinations” could gradually become mainstream “truth”.
The other potential downside, sometimes underestimated, but I need to mention – AI tools are generating unrealistic expectations that our productivity is being increased so much – so let’s trough some more work on people. Unfortunately, this is becoming reality.
But let’s see where this will lead us… There are some benefits, but there are some reasonable concerns…
So, is AI just a buzzword?
To be honest, AI does sound a bit more abstract compared with other key Data Science concepts – like Machine Learning, Neural Networks and Deep Learning – which I find more “easily” explainable – at least they sound more “mathematical” …
However, if we look a bit closer, ML is often considered as only a subset of the Artificial Intelligence, optimized for analyzing large datasets, building predictions, producing trends, making forecast etc. [2][3] – it is based on clear mathematical modeling (which does not imply “mathematic” is not a foundation for AI, as well).
Neural Networks and Deep Learning, they in fact mimic human brain operations, trying to build (among others) self-learning capabilities, by replicating human thinking process (neurons and its activations). Yet again, Deep Learning is just a subset of ML, where “deep” refers to more than three layers in the Neural Network itself… And there are really many different mathematical algorithms and models (already developed) in this domain.
Understanding all this, is AI just a buzzword?
Well, yes and no…
If we look at it from the “popular” marketing side (AI can do this, AI can do that…) – it is pretty much marketing buzzword…
But, if we observe it from the technology side and the actual use of the specific (useful) AI tools – of course this is not a buzzword, but “real” Data Science technology.
Conclusion
In this article, I am trying to demystify “AI”, but also to provide some balance view; addressing both – benefits and limitations.
The proper response would be – we need to set realistic expectations, and act accordingly. Then we can turn any technology work for us in the full capacity. Is it easy? No, but let’s try…Looks like, AI, AI, AI is not going anywhere, so we have time…
About the author…
Just to say (for the reference purposes) – beside +25 years of the IT work experience, I happen to have some certificates (and knowledge, I hope) in Data Science and AI Engineering…
More details on my LinkedIn profile: https://www.linkedin.com/in/goran-d-stevanovic-1a61925/
Acknowledgment
*) Intro photo by Steve Johnson on Unsplash
References
[1] Terminator 2: Judgment Day (1991) – IMDb
[2] AI vs. Machine Learning vs. Deep Learning vs. Neural Networks: What’s the difference? (ibm.com)
[3] Data Science vs. AI & Machine Learning | MDS@Rice
[5] Gerben Wierda on ChatGPT, Altman, and the Future (substack.com)
[7] Can AI Empower People Without Creating Chaos? – SAP Community Groups
[8] SAP Build Code | Pro-Code App Development | Developer Tools and Services
如有侵权请联系:admin#unsafe.sh