【Youth Voices】AI, Innovation, Regulation, and Labor 2.0

关注时代需要

On Times We Focus

☉ 太和智库线上英文刊物《太和观察家》2024年2月刊第41期原创文章，转载请注明出处。

☉ This article is from the February issue of TI Observer (TIO), an online publication of Taihe Institute. Please indicate the source if you hope to share this article.

☉ 点击“阅读原文”，查看本期精彩内容。（文件加载需要时间，请耐心等待）

☉ Click "Read More" to access the full text of TIO vol. 41. (It may take some time to download the PDF text.)

正文1868字，读完约需7分钟。

Wordcount: 1868. The article will take about 7 minutes to read.

---

埃文

Evan Hill

·太和青年观察家

·TI Youth Observer

Introduction

AI has become a popular concept, albeit a frequently misunderstood one, even within the IT space. How should we comprehend AI and its implications across various industries to adapt to the changing demands of the workforce and future economic development trends?

Defining "AI"

Artificial Intelligence (AI) has become a popular buzzword for a group of diverse technologies lumped together as a singular concept based on their functions. While it can be tempting to jump on the bandwagon, it is more rewarding and technically accurate to understand these individual technologies for what they truly are.

Most AI programs familiar to the public are better defined as generative large language models (LLMs) capable of sorting and outputting multiple types of natural language. LLMs are not alive, sentient, or capable of understanding inputs and outputs. They are simply highly effective word matching systems. These programs evaluate massive datasets and generate a "Frankenstein" output based upon them. They excel at replicating successful results by utilizing large sets of data rather than actually comprehending the content.

LLMs consist of two basic components - a large-scale generative system, typically based on the transformer architecture proposed by Google in 2017, and a massive curated dataset used to "teach" the LLM. The transformer architecture is well-known and accessible, but acquiring training data often poses a challenge.

While distinguishing what constitutes an AI may seem mundane, it is important to understand the underlying technical fundamentals in a market that has been caught up in hype cycles. In fact, starting from 2024, almost every startup seems to include some language about AI in their pitch deck. Although many of the new players and technologies in the LLM field are interesting and innovative, the loose definition of "AI" has allowed certain entities to repackage inferior technologies with outputs superficially similar to those produced by LLMs as AI products. 

The EU AI Act

In response to the popularization of LLMs and the presence of irresponsible actors in the field, the EU unrolled a proposed AI Act aimed at mitigating security risks associated with this emerging technology. Regulating AI is a good and necessary idea, but effective regulations require a top-down understanding of the industry, and strategic insight into how both public and private entities will react to these restrictions, including consideration of realistic enforcement options.

The General Data Protection Regulation (GDPR) is an excellent example of a poorly conceived regulation aimed in the right direction. Though the GDPR has evolved since its inception into a more effective tool, the initial act fundamentally misunderstood the objectives and technical methodologies involved. This led to near infinite exploitation of loopholes as first and third parties harvested data with impunity, using methods not banned or enforceable by the GDPR. The GDPR was eventually reworked into a much more enforceable regulation. Paradoxically, tightening the GDPR also created problems - the limitations and legislative burden of operating in Europe caused key industries to pull investment from the market.

There are valid criticisms of the AI Act, but it is worth acknowledging that it is not completely negative. Cynical Luddites and partisan geopolitics may have played a role in its creation, but the act does tackle some key issues. Certain important items, including automated educational systems, untargeted scraping of facial images, and use of LLMs in legal mechanisms, are addressed.

Unfortunately, despite some upsides, the AI Act still raises concerns. The act is fundamentally constructed by sorting broadly defined activities within the AI sphere into different risk categories, many of which could theoretically be leveraged against almost any LLM entity due to the rather broad and non-technical definitions provided. As Daniel Castro, Vice President of the Information Technology & Innovation Foundation (ITIF), stated, "Given how rapidly AI is developing, EU lawmakers should have hit pause on any legislation until they better understand what exactly it is they are regulating… Acting quickly may give the illusion of progress, but it does not guarantee success."

There are ultimately two likely outcomes for the 258-page regulation. The first assumes laxer enforcement and could resemble the ill-fated early GDPR, where entities exploited loopholes and broad definitions to do as they please. The second is a scenario where broad definitions are enforced vigorously, necessitating weaker operations and imposing significant financial and legal burdens on entities within the EU, leading to an environment where entities may decide to simply go elsewhere. The EU AI Act was built to be a sledgehammer, yet the construction of such a crude and forceful instrument may create more problems than it solves when applied to a nuanced and technical situation.

AI and the Labor Economy

Techniques and legislation aside, the question on many people's minds is how LLM technology will affect the workforce. When contemplating potential LLM impacts, it is important to note that there is a general lack of understanding in the field of AI, even amongst IT workers, which opens opportunities in itself.

Many view AIs as competitors to human labor. To some extent, this is true, but for every challenge that emerges, opportunities also arise. Automations have developed fundamentally differently than many people expected. The initial pervading belief within the automation industry was that general-purpose automatons would supplant blue-collar jobs before they began to take over white-collar work. This was very wrong in hindsight. Though currently white-collar fields are experiencing the most displacement, many traditional blue-collar roles in fields including transportation, construction, hospitality, and food and beverage should expect similar transitions as technology evolves.

The foreseeable future will be dynamic, and the rapid acquisition of new technical skills will be a new requirement for present and future generations as fields evolve. Fields will not disappear, as automations require substantial human oversight, but the workers of today and tomorrow may need to acquire different skill sets than ones that traditionally mattered.

Fields in Depth

Though the general adage of "adapting to the times and picking up technical skills" is good shorthand advice for addressing current automation-induced job displacement, there are nuances and specifics within each field. While it is difficult to predict the exact timing and trajectory for specific fields, some general trends regarding AI can be identified.

Programming is expected to be devastated by LLM adoption. Generative LLMs are surprisingly good at churning out code and will only improve further. Many low-level programming tasks can be performed more effectively and efficiently by programs like GitHub Copilot than by amateur developers. Programming will remain an important and valuable field, but the skill and specialty ceiling will rise considerably. It will evolve into a field that focuses less on writing code and more on visualized problem-solving (as many argue that high-level software engineering already does). Web developers, app builders, and programmers who implement other "solved" applications should transition to more unique areas where they can implement original solutions.

Marketing, PR, and advertising will experience tremendous disruption as the fundamental skill sets required to excel in these fields change. Traditionally, these fields have employed many copywriters and graphic designers who now no longer provide the value they once did. These fields are likely to shift into analytical and data science adjacent fields where the ability to evaluate and manage audiences is valued over the ability to personally manufacture creative materials. The modern world has long since evolved into a "choice economy," where consumers take sustenance for granted and seek comfort and self-expression, which bodes well for these fields even if the underlying skill sets change.

Journalism has been experiencing something of a crisis for a while as the industry struggles to transition from a physical medium that sustained for over a century. The current most profitable model for journalism revolves around collecting ad revenue from digital traffic, which has affected the genre of articles that make the most financial sense to publish. Ironically, the introduction of LLMs may help restore profitability in journalism, but it is more likely that we will see smaller mixed AI/human teams that are overseen by a much smaller editorial team. The industry will persist but undergo shifts as fewer people are needed.

Financial services will be disrupted tremendously as automation, digitization, and the resultant increased transparency reduce workloads. Algorithmic trading already exists, and many financial analysts are likely to evolve into roles more similar to quants than front office traders. Financial modeling and financial product design will benefit greatly from predictive algorithms that are statistically superior to human-only input. The fundamental skills needed to succeed in financial services will shift from being people-centric to technology-centric. Fields like auditing may be disrupted immensely. 

The evolution of the legal field is likely to vary greatly from country to country. It is plausible, both from a technical standpoint and based on the language contained in the EU AI Act, that at least some parts of the legal field will be disrupted by LLMs. Implementation and government controls on this are expected to be extremely tight for several security and public relations reasons, but it is plausible that some countries may be able to automate, assist with, and/or expedite certain aspects of tedious legal proceedings including contract reviews and digital filing. While it is unlikely that these changes will be evenly distributed due to technological disparities and differences in legal systems, it can be expected that some more developed nations will begin rolling out LLMs in limited support roles for some legal functions.

Blue-collar fields, although not immediately under threat, are likely to experience significant displacements in the near future. Self-driving cars and automation technologies that enable them are set to drop soon. Industries like trucking and transportation will undergo immediate changes. General purpose automatons, including fully robotic kitchens, will also be large employment displacers. Despite this, there will be a high demand for work centered around maintaining, fixing, and updating automatons in terms of hardware and software.

Conclusion

It can be easy to assume a pessimistic Luddite stance against new technology, especially when innovations disrupt the workforce. Despite these challenges, it is incumbent that a measured approach be taken when new technology arises. While LLMs have rendered some jobs obsolete, they will create different opportunities that require alternative skill sets. Like the invention of the automobile and mechanized farming equipment, for each job lost, new niches will be created. The requirements and systems change, but one thing remains constant - those who persevere and acquire meaningful skills will flourish.

The above contents only represent the views of the authors, and do not necessarily represent the views or positions of Taihe Institute.

关于TIO

太和智库线上英文刊物《太和观察家》（TI Observer）致力于促进中外沟通交流，弥合“理解鸿沟”。

TI Observer (TIO) is an online monthly English publication produced by Taihe Institute. TIO is dedicated to promoting transnational interaction and mutual understanding, thus bridging the gap of misunderstanding and bringing China and the world closer to each other.

TIO本期封面

TIO本期目录

TIO本期作者

TIO执行委员会

相关内容

【TIO 太和观察家】Rebound 2.0: Economic Recovery Amidst Global Tensions

【TIO 太和观察家】Decoupling from China: A Threat to World Economy

---

国以人兴，政以才治。

太和智库，关注时代。

长按二维码关注我们

公众号：taihezhiku

如有需要，敬请联系太和智库秘书处。

欢迎读者通过智库邮箱踊跃提交原创、首发稿件并留下联系方式，以便秘书处联系稿酬及加入智库相关事宜。

电 话：[phone]

邮 箱：[e-mail]

地 址：北京市朝阳区东三环南路顺迈金钻大厦23层

文章不错，点个“在看”吧！