Report Update: Human and Computer Deep Learning and the Future of Humanity
New Chapter on School Guidance; updates on technology, the labor markets, and deep learning
Greetings,
We’ve made many updates to our original free report, Humanity Amplified: The Fusion of Deep Learning and Human Insight to Shape the Future of Innovation. The report, which is now 246 pages with 12 chapters and 1600 endnotes, covers all the basics of what educators need to know about AI. It also recommends immediate action to develop AI literacy and human deep learning programs that are augmented by artificial intelligence to prepare students for what Bill Gates has called the “AI World.”
These are the highlights of the updates.
Guidance for schools. We added a chapter on guidance and suggested frameworks for policies for schools. The chapter reviews the guidance from UNESCO, Australia, the U.K., California, and Oregon. The review focuses on synthesizing common themes in those guidelines as well as some strengths and limitations. In the following chapter, we focus on practical steps schools can take to go through the process of developing guidance and creating a framework or policy. We focus more on frameworks because enforcing any type of AI policy in a world where technology is so ubiquitous and becoming more and more human-like will be difficult to regulate via policy.
An AI framework for your school is not something that can be decided externally, as both how we prepare students for an AI World will impact many aspects of a school’s curriculum, and what uses of AI you may choose to allow or encourage will be driven by your school’s values and overall learning goals for your students. Those vary across schools, districts, states, and countries.
Employment and training. We updated the labor markets section in Chapter 3, as AI is starting to undermine employment in certain areas. There is a growth in demand for machine learning experts, but most individuals who lose their jobs due to AI and are at risk of losing their jobs to AI generally don’t have easy paths to becoming machine learning experts.
The overall employment debate between economists and AI experts rages (it will cause massive job losses vs. it will cause some job losses and new jobs in the future). Still, I think it’s fair to say that everyone agrees that somewhere between 60 and 100% of both new and existing employees need to be trained to work with AIs (they need to learn how to work with AI tools). This is not a debate, and AI skills are currently the second most sought-after job skills (after soft skills). Figuring out how to get everyone trained to use these tools is currently a huge concern in industry, despite the financial pressures we see at many local and regional colleges. Perhaps some more alignment between what is being taught and what industry needs may help both education and industry. Right now, it seems like we have a market failure, with industry short of students who have soft skills and generative AI skills and local and regional colleges closing their doors due to a lack of students.“Concerning the education system, it should introduce more fundamental AI skills, adapt faster and work together with industries. “ Beyond Bits and Alogrithms
A summary of some suggestions for education is in this table.
There is a little more on this below the line at the bottom.
AI Literacy. We updated the case for AI literacy. Nothing can probably make it stronger than a fake CNN newscast that is making its way around social media.
Beyond this specific example, we are going to start seeing a radically different world (deep fakes, job shifting (at a minimum), the use of autonomous weapons in warfare, and potentially a complete restructuring of the entire economy).
Deep Learning. This report started as a paper that compared concepts and ideas in computer deep learning to those in academic deep learning programs, and more and more is being written about students developing the skills human deep learning programs support—critical thinking, communications, collaboration, problem-solving—and helping students become life-long learners who are aware of their learning.
The most recent Microsoft Future of Work report highlights the importance of developing these skills, and argumentation/debate (the Toulmin model) in particular.
In Chapter 4, we cover specific programs that can be implemented (in some cases, these just need to be expanded) to strengthen academic deep learning.
Technological progress. Chapter 2 is updated to focus on some recent technological developments. I don’t think it is important that educators (or anyone) follow every technological development (no one can keep up), but most leading AI scientists say that we are getting very close (or may have already arrived at) AIs that can reason and plan. Such abilities will radically expand what AI can do and have a dramatic impact on labor markets and what AI tutors can accomplish. This is just part of the process of developing human-level robots (why no one should be surprised and why only understanding the general trends matters for most people).
Debate. The more I learn about AI the more I’m committed to academic speech and debate. Debate should not just be an after-school activity, as it is in some places, but a core part of the curriculum, both as a course and integrated into all courses. Why?
(a) Debate develops the core skills and mindsets referenced in the Microsoft Future of Work report that will be needed for students to succeed in an AI World.
(b) Debate develops the communication skills needed to communicate well with AIs over the long term.
© Learning to examine an issue from two sides is a critical job skill.
See also: Bauschard, Stefan and Coverstone, Alan and Rao, P. Anand and Rao, Sebastian, Beyond Algorithmic Solutions: The Significance of Academic Debate for Learning Assessment and Skill Cultivation in the AI World (September 4, 2023). Available at SSRN: https://ssrn.com/abstract=4567346 or http://dx.doi.org/10.2139/ssrn.4567346
___
From Beyond Bits and Alogrithms
Most companies expressed the increased importance of certain hard and soft skills. Some stated that soft skills are “much more important than the hard skills” (HR-11) and that “when you are approaching teamwork /…/ the importance of soft knowledge is rising. And also if you are trying to promote new technologies to your employees, you need to have soft skills” (HR-6). Companies highlighted the importance of willingness to learn, agility, good communication, collaboration, active listening, analytical thinking, resilience, innovation and creativity, strategic and logical thinking, as well as problem-solving. “One additional skill set which is also very important is critical thinking. And this part is usually quite challenging because here you should not blame, but you should critique” (HR-11). Others did not necessarily pinpoint critical thinking, but most of them are aware that “you cannot be reliant on artificial intelligence. It is not 100 percent bulletproof” (HR-4), as it is necessary to evaluate the outputs of AI systems and ensure their accuracy and reliability. Furthermore, when using generative AI, the skill of prompt engineering is also important. “You always have to ask the right question to get an explanation on how and on what the decision was made” (HR-9). “Right now, they are saying a lot of positions are closing, and they just need a position for prompting” (HR-5). However, despite the ever-emphasised importance of soft skills, hard skills are becoming a hot topic regarding AI implementation. “I believe the combination of hard and soft skills is probably the best” (HR-12). “Hard skills, I would say, will always be needed. It only depends on how many people we still need with good knowledge of these hard skills and what hard skills we really need” (HR-2). Hard skills that were mentioned were programming (Python, R, SQL), data management and data analysis, knowledge of specific AI tools and machine learning algorithms, but also more fundamental knowledge of mathematics, logic, and statistics. “With AI technologies, it is hard to find individuals who can be multidisciplinary and have all the necessary skills” (HR-19). The majority also mentioned the significance of knowing the business processes themselves. An important difference between the companies is to what extent they expect their employees to dive into the learning of programming. “Basic knowledge should be a necessity for leaders. Not everyone should learn how to code, especially on lower levels where people do not need to make decisions” (HR-18). Most companies are looking for people to fill the positions that will also be in contact with AI. “Currently, there are no suitable candidates with all the necessary skills in the market. Specific knowledge is required /…/ both in terms of understanding certain models and business processes and the data they generate” (HR-15). This is why companies are not looking for people — 239 — who have a full understanding of AI but rather those who are willing to learn and grow as AI develops further. Moreover, “today the demand for this kind of [AI] solutions and consequently people who are able to build these kinds of solutions is increasing. /…/ Therefore, we are looking to reskill and upskill existing employees to move to a new direction of development” (HR-11). It is also vital for companies to acknowledge the benefits of higher education for the sole purpose of successfully implementing AI models. “When comparing education, however, there are big differences between having a PhD and other degrees. AI is a field where you need a research approach with a high level of abstraction, and PhD forces you to think on the level of theory” (HR-18).
Lastly, the companies shared their opinion on the role of the state and universities in AI implementation and the necessary skills development, and their opinion was largely confirmed by the literature. The government’s actions are, Table 3. Summary of key findings Key highlights General findings Importance of learning culture, gradual transition to using AI. Focus on employees who are willing to learn and grow with AI. Soft skills needed Willingness to learn, agility, good communication, collaboration, active listening, analytical thinking, resilience, innovation, creativity, strategic and logical thinking, problem solving, curiosity and critical thinking. Hard skills needed Fundamental skills needed Mathematics, logic, statistics. Other hard skills Programming, data management, AI tools, machine learning algorithms, understanding of processes, prompt engineering. Educational methods Learning by doing, mentoring, job shadowing, coaching, online courses, workshops, knowledge sharing, and external certifications. Personalisation of goals and educational processes, an array of methods. Obstacles Mindset, time, willingness to learn, age, fear. Macro situation Need for government’s action, agility of the education system, collaboration of industries and the educational system. Source: Own work (2023). — 243 — in many cases, perceived as insufficient. It should be more active in creating large-scale programs to tackle the skills gap. Concerning the education system, it should introduce more fundamental AI skills, adapt faster and work together with industries. The key findings of the empirical research can be found in Table 3.