Maybe you just don’t have time to keep up with the advent of proxy AI. Or, you simply don’t realize that AI is not just a fashion for higher education. Instead, our greatest hope is to survive the widespread budget cuts and lower enrollment. Or, perhaps based on your opinion of AI tools in 2023, you find them too unreliable and affected by hallucinations to take them seriously. Hopefully this summer gives you the time to catch up with technology and start the fall semester with knowledge and experience to make AI the best professional assistant you have ever had.
You will have a PhD, rather than just facing a bunch of projects and problems, but instead a knowledge virtual assistant to work beyond human thought:
“According to OpenAI, O3 scored a record score on the ARC-AGI visual reasoning benchmark, reaching 87.5% on high computational tests, matching human performance at a threshold of 85% compared to human performance. The model also received 96.7% of the times in the 2024 US Invitational Mathematics Exam in 2024, on just one question, but on a dia of dia conding of dia condernd conding of dia in conted of dia.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7.7
These assistants work 24-7 and have no holidays or holiday breaks. At the end of a long work, you can ask complex questions for advanced deep research models and do research while you have dinner, write detailed reports and prepare follow-up questions (I have experienced these myself; I have experienced these myself; reading about desserts is more satisfying). I also woke up in the early hours of Sunday morning with a good idea to pursue work. I took it directly into a high-level model and woke up a report of over 20 pages and conducted a comprehensive citation and recommendation later that morning for further research.
AI developers have made great strides in avoiding hallucinations and detached results. Among them is the utilization of retrieval generation, a natural language processing technology that can leverage a spacious database to “enhance the context and accuracy of generated text”. Those who may have stopped using AI because of past mistakes will be surprised to see more consistent and accurate results today. Despite these improvements, I continue to recommend that users submit nearly the same hints to the two leading models of AI. Although the advice I made is mainly that you may get other useful information from the added results.
Today, most leading models offer multi-model capabilities that can input, process and output various types of data beyond text, such as images, audio and video. This makes it possible to attract job seekers through voice communication. It supports a generation of amazing images and fairly short video clips. Google’s notebook LM tool accepts input documents and related media to create podcasts, allows listeners to ask questions and get audio responses, create crosswords based on your handout instructions, and even create virtual debates.
The tools I use most are from OpenAI, especially O3; people from Google, especially Gemini 2.0 Flash for general work, Gemini 2.5 Pro for detailed research; and Claude 3.7 sonnets for humans. However, there are more models available today that better meet your needs or preferences. One of the projects I’ve been pursuing lately is tracking research, new therapies and other information about specific diseases. I’m using Gemini 2.5 Pro and Chatgpt O3, running updates every week. The result is comprehensive, cited. It is worth noting that Chatgpt O3 noticed that I was asking for a report every Tuesday and asked if I wanted it to run the same prompt every Tuesday and emailed the results to me. Such actions can be considered as basic proxy AI, where the tool can analyze requirements, create action plans, and take autonomous actions with permissions:
“Proxy AI is an AI system that automatically works, adapts in real time and solves multi-step problems based on context and goals. They are built by multiple AI agents that utilize large language models (LLMS) and complex reasoning. This allows them to enhance decision-making capabilities and natural language understanding, facilitating effective and intuitive user interaction.”
The power of AI agents has just begun to be realized. 2025 is called the agent year. Millions of agencies are expected to be created by the end of this year. Their potential is enormous, beyond the individual acting on behalf of human beings.
So, how can we master this AI trend this summer? Timely engineering – asking questions with the right context, details and format – is a good starting point. You may consider taking one of many low-cost or free timely engineering online courses. Google prompts Essentials courses for $49 via Coursera. This takes only a few hours to successfully complete the certificate. There are also many YouTube videos that condense micro-instrumental content. You may need to browse the Coursera Catalog section on timely engineering, which lists dozens of courses from industry leaders, business vendors, universities and universities that last for weeks to months. Many people offer professional certificates.
I suggest you start a course – many people are self-paced, or an informal YouTube video and then start using the tools whenever possible opportunities. The iteration prompt is the name of the game. Try remarking the prompts, providing additional information, and including what you are seeking the tool to better understand your expectations.
Use as many tools as you choose. Ignore non-AI search tools for a period of time. You’ll notice that AI search is the right best solution, rather than listing the response first by those who pay in search responses, just like asking a question with Google.
Search topics appear four to five times a day, perfecting each search tip to better understand the capabilities of each tool, and searching on a variety of topics and disciplines will improve your expertise and comfort in AI. By the time courses that begin in the fall, you will be ready to save a lot of time and effort by using AI. You can also integrate AI into your daily work to become more productive and share your expertise with colleagues and students.
