Gemini

Gemini is an exceptional tool in assisting and aiding work tasks especially as a SCADA Engineer due to the following: 1. Review of code and logic for implementation - In our day to day activity, it is necessary for us to make sure that before we put something in production, it is validated and tested. One of our error spotters is Gemini and it spots the tiniest errors that sometimes you cannot see in reviews as fast as this AI can do. 2. Documentation - Gemini helps in proof reading the documentation that we put out and ensures that we can validate them through other available sources that may have produced the same tests or observations. 3. Integration of Notebook LLM - Due to the number of devices we operate and maintain, we also handle a significant volume of technical documentation from various vendors and manufacturers. This includes all the technical manuals, guides and troubleshooting docs. With the integration of Notebook LLM and Gemini, we can consolidate these into a single source, searchable engine where we can ask specific and context-based questions without us having to rely on our knowledge and/or scanning numerous pages. We can also cross reference information and find ideas that might work that are not naturally written in these pages. As such, our response time is improving in diagnosing and finding solutions.

I love the polished feature and how easily it is integrated into Google Workspace. Most things are good, but I wish it would tailor things to me and not just generic

- The native integration with all the google workspace. It saves me time from searching for docs and reading through them instead I can have gemini summerize or bring it to me. - I also value the 'Thinking mode' which provides me deeper, more logical responses for complex coding and math problems. - The usage of gemini UI is very smooth and not laggy which is something I keep finding most LLMs when working with very large context window and complex problem. But Gemini is best in my experience and it is alwyas smooth and fast even for larger complex models.

In my experiences, in terms of dislike, I do not necessarily dislike the product or part of its services. There are just areas where efforts must be made in order to ensure that you get the correct data/information you need from Gemini. 1. Prompting - If you just normally ask Gemini in a way that you would ask a person, there are instances where it gives vague or off context answers. It is important for the ones who are asking the AI to perform a specific task to be specific, contextual and structured to provide clarity on the intent of the task. That way, you can get the result to as close as you would want it to be, but of course, it is still imperative that you check the results, especially if they handle factual information. 2. With its image generation, when requesting small changes to a generated image, the AI may alter the entire image instead of applying precise edits. This can result in distortions and loss of original details and output that deviates from the intended revision. Sometimes fine tuning requires us to restart Gemini and include the adjustment in the initial prompt. I think these two are the most common challenges I encounter. However, I've learned to work around them to use the AI model as efficiently and accurately as possible while adhering to the existing guidelines and policies set by our IT team.

I just think it is very basic, I need to learn more. But also I think that it needs to learn to be smarter, start tracking moves not just generic AI. What can I improve on or do differently

- Inconsistent context memory can be a bit annoying at times. - It tends to forget a few prompts after 5-6 conversations later if all the conversations are of very large context in the first place. - Some safety features can be overly sensitive, occatinally refusing to answer benign questions because they touch sensitive topics, even though it would make sense in a professional context.