Join our daily and weekly newspapers for exclusive content on the latest updates and industry-composure AI coverage. learn more
Question: Which product should use machine learning (ML)?
Project manager answer: Yes.
On one hand, the advent of jokes, generous AI has increased our understanding whether use cases lend the best for ML. Historically, we have always taken advantage of ML for the future stating pattern, but now, it is possible to take advantage of a form of ML even without the entire training dataset.
However, the answer to the question “What does the customer need, AI needs a solution?” Still not always “yes”. Large language models (LLMS) can still be prohibited for something, and with all ML models, LLMs are not always accurate. Such cases will always be used where taking advantage of ML implementation is not the right way. How do AI project manager evaluate our customers’ requirements for AI implementation?
Important ideas to help make this decision include:
- Input and output required to meet your customer needs: An input is provided by the customer to your product and the output is provided by your product. Therefore, for a spotify ml-generated playlist (an output), input may include customer preferences, and ‘liked’ songs, artists and music style.
- Combination of input and output: Customer requirements may vary on the basis that they want the same or different output for the same or different input. The more orderly and combination we need to repeat for input and output, on the scale, we need to turn to ML vs. rules-based systems.
- Pattern in input and output: Pattern in the required combinations of input or output helps you decide what type of ML model you need to use for implementation. If there are patterns of combination of input and output (such as reviewing the customer’s anecdotes to achieve an emotion score), consider the supervised or semi-revised ML model on the lLM as they can be more cost effective.
- Cost and accurate: LLM calls are not always cheap on the scale and outputs are not always accurate/accurate, despite fine-tuning and early engineering. Sometimes, you can classify an input using a certain set of labels instead of using an LLM that is better with a supervised model to the nerve network.
I put a quick table together below, summarizing the above views, to help the project managers evaluate their customer’s needs and to determine whether the ML implementation sounds like the right path.
| Customer needs | Example | ML implementation (yes/no/depends) | Type of mL implementation |
|---|---|---|---|
| Reaping | Add my email to various forms online | No | Creating a rule-based system is more than enough to help you in your output |
| Reaping | The customer is in “Discovery Mode” and expects a new experience when they take the same action (such as signing into an account): – Click a new artwork per click ,Stumbling (Remember?) Search for a new corner of the Internet through random discovery | Yes | -Mage Generation LLM -ReCommendation algorithm (collaborative filtering) |
| Reaping | Receive essays Customer response to increase subjects | Depends on | If the number of input and output combinations is quite simple, a determinable, rule-based system can still work for you. However, if you start having many combinations of input and output because a rule-based system cannot be effectively on scale, consider bending: -Clacifier But only if there are patterns of these inputs. If there is no pattern, consider taking advantage of LLMS, but only for one-closed scenarios (as LLMs are not as precise as supervised models). |
| Rejected output | Customer aid questions complete -Search | Yes | This is rare to get into examples where you can provide different outputs for different inputs on a scale without ML. There are many orderly orderly order for a rule-based implementation. Consider: -Lams with Retriel-Agmented Generation (RAG) |
| Non-dear work with separate output | A hotel/restaurant review | Yes | Pre-LMS, this type of landscape was difficult to complete without models who were trained for specific tasks, such as: -Curant Neural Network (RNN) LLMS are a great fit for this type of landscape. |
Bottom line: When a simple pair of scissors could trick, do not use lightsber. Evaluate your customer’s requirement using the top matrix, taking into account the cost of implementation and the accuracy of the output, to manufacture the precision, cost -effective products on the scale.
Sharanya Rao is a Fintech Group Product Manager. The views expressed in this article are of the author and not necessarily the people of his company or organization.,
