Declutter AI: things that matter!
Factors that drive the economics of AI projects are markedly different.
If you are reading this, chances are that you are evaluating an AI-driven tech project and grappling with questions about the economics of your project.
While the ROI equations are fundamentally the same, the factors that drive the costs and the benefits of AI are markedly different and unique. So, what are these? Here is a practitioner’s perspective at a micro level. In this series, we will also cover the macro-economic level as well as the ethical and ESG aspects of AI.
Microsoft has introduced advanced AI solutions aimed at streamlining tasks and automating repetitive processes. Microsoft's AI-powered Copilot accelerates customer response times, optimizes scheduling, and facilitates task recommendations. This innovation addresses challenges posed by labor shortages and skill gaps among the global frontline workforces.
Watch Shantanu Narayen talk about how he is reimagining Adobe in the age of AI, the use of generative AI in the creative industry to automate content production, and touches upon how it will augment human intelligence and not replace it. He mentions three stages in which Adobe is using data to develop models, collect data, and build interfaces on Adobe applications.
Ikigai Labs, secured a $25 million investment led by Premji Invest to advance their generative AI platform for tabular data. Ikigai Labs empowers enterprises with AI-driven solutions addressing challenges in data usage and scenario-based planning leading to efficient forecasting and analysis capabilities. The funding totals $38.2 million and will support the expansion of Ikigai's team from 30 to 70 employees.
The Indian tax system, with a base of above 70 million registered taxpayers, closed its recent filing cycle on July 31st. This is also an example of how AI and Data Analytics have been leveraged at a scale to widen the tax base and make the tax filing processes a better experience for the citizens.
Widen the tax base - A record 67.7 million income tax returns were filed for the assessment year (AY) 2023-24, and this includes 5.3 million who filed their returns for the time. The number of taxpayers earning over Rs 1 crore (US $120K) has seen a 51% increase since 2014-15.
Improved intelligence on income and taxes, the wider application of tax collection at source, and better reporting by banks have contributed to this change.?As the technology gets deeper and wider, the number of filings and tax collections should keep increasing year after year.
Tax Filing efficiency - Technology has also been leveraged to simplify the compliance burden on taxpayers, and to identify tax evasions. Pre-filled ITR forms, faceless assessments, and appeals to remove the physical interaction with the IT department have eased the income tax filing process. Public queries and grievance redressal are handled through new-age systems - Centralized Public Grievance Redress and Monitoring System (CPGRAMS), e-Nivaran, and Aaykar Seva Kendras (ASK). As a result, during the tax filing month of July, the 2-crore milestone was reached 9 days earlier this year.
Detect tax evasion - AI-driven processes cross-reference the transaction basis the identifiers of the transacting party (the Aadhaar and the PAN number), and further create linkages with the bank accounts and therefore able to compile comprehensive information regarding fixed deposits, quarterly interest credits, share dividends, share transactions, mutual fund investments, long-term capital gains, short-term gains, etc. and flag the discrepancies.
ChatGPT is not replacing human intelligence, not yet!
In the last few months ChatGPT in particular, and LLM in general, raised the interest levels in AI to unprecedented levels. Never were regulators, business operators, and laymen alike, so interested in the possibilities of AI. Human beings of all ilk raced to “write” poetry, stories, jokes, and even complete novels. Dozens of times ChatGPT made news because its servers crashed trying to handle all the traffic coming in. Recently, the usage has reduced. Many believe that competition from Google's Bard, and others too, will bring sanity into the Generative AI market. This is great news because when the hype fritters away, the real business starts.
And then the big wave started seeing some real breakers. In July, The Washington Post reported a net 10% reduction in traffic to ChatGPT.
They are not trained to tell people they don’t know what they’re doing.” said Yilun Du, a researcher at MIT who was previously a research fellow at OpenAI. The result is bots that act like precocious people-pleasers, making up answers instead of admitting they simply don’t know. (The Washington Post, May 23, 2023).
Users report that generative chatbots have been “making up” false information. Worse, most do not provide any trail on how they determined the information being handed out, not even the sources referred.
So, our data analyst team got very interested in the question.
Where can ChatGPT actually help?
And the outcome is that we discovered the boundaries of it being useful and reliable. It's great to help with tasks that might otherwise be routine and mundane. But when it comes to applying the context of the systems and the processes to identify the right solution, one cannot replace human intelligence and ingenuity. We relate to it because in our real-life project implementations, we as techies, always work with business experts who have the contextual knowledge of the problem we are trying to solve.
Helps to reduce effort and time with certain routine tasks.?
For example, the generation of SQL queries. This is great because it frees up the tech bandwidth which is very expensive. As an example, below, the analyst needed an SQL query to join orders and tables, based on order ID, filtered on country, and a count of orders city-wise and aggregated profits city-wise.
And there are quite a few other examples. Code improvements by subjecting your code to reviews or debugging, generating documents, or the SQL syntax to name a few. And there is a recognition, to get the best out of ChatGPT, or any LLM, one needs the very human talent of prompt engineering. More on that in the next edition!