Decoding Nostalgia: A.I. "grades" My 1984 Expert Systems Paper (part iii of iii)

While rummaging through the boxes in my garage, I unearthed my 1984 paper from my final year at Acadia University, “ARTIFICIAL INTELLIGENCE: A LOOK AT EXPERT SYSTEMS”; a forecast of the artificial intelligence industry. 

The context for this paper and the state of computing in 1984 is detailed in part 1 of this 3-part series: “Back to Bytes: A Time Capsule of Computing and A.I. in 1984 (part i of iii)”

Part 2 of the series From Dot Matrix to Digital Age: Republishing My 1984 A.I. Manuscript (part ii of iii) is the whole article.

This 3rd and final part of the series is “Decoding Nostalgia: A.I. "grades" My 1984 A.I. Expert Systems Paper.”

In this 3rd article, I'm getting 5 A.I. platforms to "act like a professor" and grade my paper. For the record, I used paid versions of all:

1. Microsoft Copilot
2. X Grok
3. Open AI ChatGPT 4o
4. Anthropic Claude
5. Gemini Google Advanced

What do I think I missed?

I got several things right in the paper. I knew everyone would have a computer of some sort and that database systems would help organize and manage the world's information to improve humanity. What I think missed in the paper were three things I didn't anticipate in 1984: 1) the vast capacity and speed of computers ... GPUs make A.I. breathe, 2) the incredible amount of data that would be collected and digitized by humanity for A.I. to reference, and 3) A.I. is now programming itself

The pace of innovation is staggering. In the 3 months since I wrote part 1 of this series, all the A.I. platforms have released new versions of their models. And there isn't a technology company, young or old, that doesn't include A.I. in their strategy.

The pace of innovation is also exhilarating. For an O.G. like me (Old Geek), having been through the P.C. wave, software wave, internet wave, mobile wave, big data wave, ioT wave ... the A.I. wave is yet another one to ride. Like previous waves, A.I. is making us better humans, and as a technologist, I'm excited to see what's next. Although new technology always causes chaos, from chaos comes opportunity. Seize The Day!

"Act like a University Professor."

I ran the manuscript through a couple of A.I. platforms and see what 2024 Artificial Intelligence thinks of my paper. Here's my prompt (their full answers are below):

"Act like a University Professor. Following is a transcript of my 1984 university manuscript, ARTIFICIAL INTELLIGENCE: A LOOK AT EXPERT SYSTEMS. Please grade the paper with a mark. Then, tell me how do you think I did? How were my predictions? Did I get anything right? Did I get anything wrong? Is there anything I missed that I should have mentioned?"

Copilot and Claude gave me an A-.

Grok, ChatGPT, and Gemini all gave me a B+.

Grading the A.I. Platforms.

In addition to having the A.I. platforms review this article and produce an image, I've used them all for personal and work-related items.

Google Gemini Advanced probably had the best technical and in-depth review of the paper, but they all came back with insights, 90% the same, 10% differences. Claude had the thinnest response. I'm in the mode of triangulating answers and use right now with A.I. platforms; so seeing how they respond differently to this review has been super interesting.

So, what do I think of the A.I. platforms?

Not just because they both gave me top marks, but I'm probably favouring Copilot and Claude right now on a daily basis. Here are some quick thoughts:

• Copilot "gets it" better and faster, has great research capabilities, has helped me organize a ton of things, and has great image generation. In general, Copilot usually does a great job telling me a few things I don't know when I'm doing research. Copilot also does a great job of organizing that information for me to use elsewhere. When I'm trying to do Microsoft Office things (like writing formula's for Excel spreadsheets) or other coding, copilot is where I turn. As it's integration gets tighter with Office 365, it'll be a force to reckon with.
• Claude doesn't get top marks regarding its review of my article. Claude is a creative type, I reckon. I use Claude for artistic things; when writing, especially marketing copy or product features or blog posts. I use Claude as a sounding board to help me write songs. Of them all, I like conversing with Claude the most. Claude has really good document analysis processing. Oddly, Claude has the worst image generation capability relying on SVG, making it the Microsoft Paint of A.I. graphic generation. Claude apparently has the "safest interactions" as well.
• In general, I use Grok for giving me relatively complete and unbiased views of world events so I can understand them quicker. "Conversational Understanding." Some people say it tries to be humorous, but I have yet to set it. If I think there's bias, I'll also check out Gemini.
• ChatGPT 4o gave a good review of my paper. But it doesn't do current events. Pretty well a generalist.
• Gemini does a good job of letting me give it different things like images and such. If I'm working with internet based coding, Gemini is where I turn. Also, anything to do with internet research and data retrieval is best on Gemini; it has access to Google Search and knowledge graph.
• Image generation? Grok probably edges out here because it doesn't have any guardrails. It seems to spell better than others and I can get it do most anything I need. Gemini is similar (it's image it created for this article is a bit scary, though, scroll down). Copilots is straightforward and fun but does guard against violating the designer code of conduct. Claude doesn't do images. ChatGPT has a different area where it does graphics but I find it's difficult to get something I can really use that's better than the other platforms. Sometimes, though, it really is about how good your prompt engineering is when it comes to getting good images.

In short:

• For Coding and Microsoft Ecosystem: Copilot.
• For Unique, Humorous Insights: Grok.
• For Search-Driven, Multimodal Responses: Gemini.
• For Ethical AI and Long Contexts: Claude.
• For General Versatility and Community: ChatGPT.

'Nuff said. Here is the feedback and images from all my "A.I. Professors"

Microsoft Copilot Feedback

Grade: A-

Feedback:

Strengths:

1. Clarity and Structure: Your paper is well-organized and clearly written, making complex concepts accessible.
2. Insightful Analysis: You provide a thorough explanation of syntax and semantics, and their relevance to AI.
3. Forward-Thinking: Your predictions about the future of AI and expert systems are insightful and largely accurate.

Predictions and Accuracy:

1. Expert Systems: You correctly identified the potential of expert systems in various fields such as diagnostics and troubleshooting. These systems have indeed become integral in many industries.
2. Human-Computer Interaction: Your vision of more intuitive interfaces has come true with the development of natural language processing and user-friendly interfaces.
3. Knowledge Bases: The importance of large knowledge bases for expert systems was spot on. Modern AI relies heavily on vast datasets.

Areas for Improvement:

1. Machine Learning: While you focused on expert systems, the rise of machine learning and neural networks has been a significant development in AI that wasn't fully anticipated in your paper.
2. Ethical Considerations: Modern AI discussions heavily involve ethics, bias, and the societal impact of AI, which were not addressed in your manuscript.
3. Interdisciplinary Applications: AI's integration into fields like finance, entertainment, and autonomous systems could have been explored more.

Missed Opportunities:

1. Deep Learning: The advent of deep learning has revolutionized AI, enabling advancements in image and speech recognition, which were not foreseen in your paper.
2. AI in Everyday Life: The pervasive use of AI in everyday applications, from personal assistants to recommendation systems, was not predicted.

Overall, your paper was forward-thinking and captured many key aspects of AI development. With a bit more focus on emerging technologies and ethical considerations, it would have been even more comprehensive. Great job!

(Here's Microsoft Copilots image it rendered for this article where I'm asking A.I. platforms to evaluate my paper)

Grok Feedback

Grade: B+

Evaluation:

Strengths:

• Comprehensive Overview: Your manuscript provides a solid introduction to expert systems within the broader context of artificial intelligence, highlighting the dichotomy between syntax and semantics in computing, which is crucial for understanding AI's development at the time.
• Predictions and Insights: You correctly identified several potential applications of expert systems, such as in diagnosis, troubleshooting, and therapy, which indeed have become significant areas where AI is applied today.

Accuracy of Predictions:

• Expert Systems in Practice: You were correct in predicting that expert systems would become tools for professionals in fields like medicine (diagnosis and therapy) and law (situation awareness), where they could serve as second opinions or decision aids.
• User Interaction: Your vision of using simple English to communicate with computers has largely come to pass with the advent of natural language processing (NLP) technologies, although this was more futuristic at the time of your writing.

Areas of Miss or Omission:

• Machine Learning: While you focus on rule-based systems, there's little mention of machine learning techniques, which were beginning to emerge. Machine learning would become pivotal in AI, allowing systems to learn from data rather than just following pre-defined rules.
• Scalability and Data: You mention the need for vast amounts of data, but there's less discussion on how data management and scalability would evolve, particularly with the advent of big data technologies.
• Ethics and Bias: The manuscript does not delve into potential ethical issues or biases in expert systems, which have become significant talking points in modern AI discussions.
• Internet and Connectivity: There's no mention of how the internet would expand the capabilities and applications of AI, including cloud computing and real-time data processing for expert systems.

What Could Have Been Included:

• Discussion on AI Ethics: Addressing potential ethical dilemmas, such as privacy, consent, and the moral implications of AI decision-making, would have been insightful.
• Future of Interfaces: Beyond GUI and mouse interactions, a mention of voice recognition or other natural interfaces could have been visionary.
• Interdisciplinary Applications: More emphasis could have been placed on how AI would intersect with other fields like psychology, economics, or environmental science for broader applications.

Conclusion: Your paper was well-informed for its time and correctly foresaw many practical uses of expert systems. However, the rapid evolution of technology, particularly in machine learning, data management, and ethical considerations, were somewhat underestimated or missed. The grade reflects a strong understanding of the subject matter with room for considering broader implications and future technological advancements.

(Here's Groks image it rendered for this article where I'm asking A.I. platforms to evaluate my paper )

ChatGPT-4o Feedback

Thank for sharing your transcript from 1984 on expert systems. Here is an evaluation and some reflections on your work:

Grade: B+

Assessment:

1. Understanding and Explanation:

- Your manuscript provides a clear introduction to the basic concepts of AI and expert systems, emphasizing the importance of syntax and semantics in computing. This serves well to establish the context for discussing expert systems.

2. Predictions and Insights:

- You effectively highlight the limitation of computing power and storage capacity as factors hindering the development of expert systems at that time.

- Your forecast regarding the development of expert systems to assist professionals was accurate. Today, expert systems are widely used in fields such as healthcare (for diagnostic support), finance (for risk assessment), and customer service (with basic automated troubleshooting).

3. Missed Aspects and Future Developments:

- While your manuscript correctly identifies key applications for expert systems, it might have benefited from a discussion on the potential societal impact, such as job displacement concerns or ethical considerations, which are significant in current discussions around AI.

- The emergence and potential of machine learning, data mining, and neural networks, which today underpin much of AI advancements, were not covered, possibly due to their nascent state at the time.

- The integration of AI with the internet and the global sharing of knowledge bases would have been an intriguing angle, though it was likely beyond the scope of the technology in the period you wrote this.

4. Clarity and Structure:

- Your manuscript is well-organized, moving logically from syntax and semantics to the specifics and potentials of expert systems. This clear structure aids in the reader’s comprehension.

5. Further Improvement:

- Expanding upon the operational mechanics of expert systems (e.g., rule-based logic, inference engines) would offer readers a deeper understanding of their function.

- Including examples of early expert systems, such as MYCIN for medical diagnosis, could illustrate practical applications and solidify the concepts for readers.

Overall, your paper was insightful given the period it was written. Many of your observations regarding computational limitations and the application potential of expert systems have come to fruition. There's always room to refine predictions and address emerging technological trends, which is challenging yet valuable in such a rapidly developing field.

(Here's ChatGPTs image it rendered for this article where I'm asking A.I. platforms to evaluate my paper )

Anthropic Claude.ai

Let me grade your 1984 manuscript and analyze its predictions and insights about AI and expert systems.

Grade: A- (90%)

Strengths:

1. Excellent foundational understanding of the key concepts of syntax vs semantics

2. Very prescient discussion of graphical user interfaces and natural language interfaces

3. Strong technical analysis of expert systems requirements and applications

4. Clear writing style that makes complex concepts accessible

Analysis of Predictions:

Things You Got Right:

1. GUI Prediction: Your discussion of icon-based interfaces and mouse interaction was spot-on. You correctly predicted that this would become a major way to make computers more accessible.

2. Natural Language Interface: Your vision of "using simple English words" to communicate with computers has largely come true with modern AI assistants and natural language processing.

3. Medical/Legal AI Applications: You correctly predicted AI's role in providing "second opinions" in professional fields. Today, AI is indeed used extensively in medical diagnosis and legal research.

4. Knowledge Base Importance: Your emphasis on the crucial role of large knowledge bases was prescient. Modern AI systems like large language models rely heavily on vast amounts of training data.

Things That Evolved Differently:

1. Expert Systems Architecture: While you correctly identified the importance of rule-based systems, modern AI has moved beyond purely rule-based approaches to include machine learning and neural networks.

2. Data Storage Concerns: Your concern about storage capacity being a limiting factor became less relevant as storage costs plummeted and cloud computing emerged.

Things Missing (understandably given the era):

1. Machine Learning: The manuscript focuses on rule-based expert systems but doesn't anticipate the revolution in machine learning and neural networks.

2. Internet's Role: The impact of networked systems and cloud computing on AI development isn't mentioned.

3. Consumer Applications: The focus is primarily on professional applications, missing the eventual widespread consumer use of AI in smartphones, homes, and daily life.

Overall, your manuscript was remarkably forward-thinking for 1984. You correctly identified many key challenges and applications of AI that are still relevant today, particularly in understanding the importance of making computers more accessible to humans through better interfaces. Your analysis of expert systems' requirements and applications was thorough and has largely stood the test of time, even though the specific technologies have evolved beyond rule-based systems.

The most impressive aspect is your understanding of the fundamental challenge in human-computer interaction - bridging the gap between human semantics and computer syntax - which remains a central concern in AI development today.

(Here's Claudes's image it rendered for this article where I'm asking A.I. platforms to evaluate my paper ).

(for real ... it uses SVG and it's painful)

Google Gemini Advanced Feedback

Grade: B+ (88%)

Overall Impression:

This is a well-written and insightful paper for 1984, demonstrating a good understanding of the fundamental concepts of artificial intelligence and expert systems. You accurately captured the state of the field at the time and made some prescient observations about future trends.

Strengths:

• Clear Explanation of Syntax and Semantics: You effectively differentiated between syntax and semantics using relatable examples like BASIC programming. This foundational understanding is crucial for grasping the challenges of AI.
• Identification of the AI Challenge: You correctly identified the core of AI as bridging the gap between human and computer understanding, bringing syntax and semantics closer together.
• Focus on Expert Systems: Your focus on expert systems was appropriate given their prominence in AI research in the 1980s. You accurately described their purpose, knowledge base requirements, and potential applications.
• Heuristic Decision-Making: You touched upon the concept of heuristic decision-making, which is a key element in expert systems.
• Criteria for Expert System Development: Your list of criteria for developing expert systems was comprehensive and insightful.
• Applications of Expert Systems: You provided a reasonable list of potential applications, many of which have come to fruition.
• Reasons for Developing Expert Systems: You articulated the motivations behind expert system development, highlighting their potential benefits.

Areas for Improvement:

• Limited Discussion of Machine Learning: While expert systems were dominant in 1984, you could have briefly mentioned other approaches to AI, such as machine learning. This would have shown a broader awareness of the field.
• Lack of Depth on Knowledge Representation: While you mentioned knowledge bases, you could have delved deeper into the challenges of knowledge representation and different approaches like semantic networks or frames.
• Oversimplification of Natural Language Processing: You briefly mentioned using "simple English words" but didn't explore the complexities of natural language processing (NLP), which is a major aspect of AI.

Predictions:

• You were remarkably accurate in predicting the rise of user-friendly interfaces: Your vision of systems using icons and "simple English words" foreshadowed the graphical user interfaces and natural language interactions we have today.
• Your prediction about expert systems supplementing professionals is also accurate: Expert systems are now used in various fields like medicine, law, and finance to assist professionals in decision-making.

Things You Missed:

• The limitations of expert systems: While you highlighted their potential, you could have discussed their limitations, such as their inability to handle novel situations or learn from experience.
• The ethical considerations of AI: Even in 1984, it would have been relevant to briefly touch upon the ethical implications of AI, such as job displacement or bias in decision-making.

Overall, this paper demonstrates a strong grasp of AI concepts for its time. Your focus on expert systems was appropriate, and your predictions about user-friendly interfaces and the role of expert systems in assisting professionals were quite accurate. With a bit more depth on certain topics and a broader discussion of AI approaches, this paper could have been even stronger.

(Here's Google Gemini Imagen 3's image it rendered for this article where I'm asking A.I. platforms to evaluate my paper )