Gig Economy, AI, and LLM: Friends or Foes?

Fiverr, one of the marquee players of the online gig economy, announced several product improvements earlier this week. Fiverr is integrating AI to enhance business access to freelancers. On Tuesday, Jan. 30, the platform unveiled an AI-powered homepage, a key feature in its new product offerings:

“‘Fiverr’s transformed homepage offers more personalization, seamless navigation, and additional control, simplifying access to the right freelance talent for businesses,’ the company said in a news release provided to PYMNTS. ‘Leveraging “Fiverr Neo,” Fiverr’s AI-powered copilot technology, businesses will get tailored service recommendations to enhance their project,’ the company added.”

This is interesting on so many levels, but let’s take it one step at a time. 

Looking at Fiverr’s stock performance since its IPO, it’s evident that the firm has lost more than 90% of its value. And while there are ample reasons for this, with a focus on SMBs (which tend to be more transactional) and macro-related pressure, part of it is driven by the fear that many of the services offered via Fiverr can be done better and cheaper using ChatGPT or Midjourney. For example, I used Fiverr to create graphics, and now I only use Midjourney or DALL-E (which created the image for this article). And for audio transcription services, I now use Otter instead of the human option provided by Fiverr.

But the real question is whether these online gig economy firms have any future in a world of Gen-AI firms. 

Introduction

The gig economy is an economic model largely driven by companies operating through digital platforms, creating a marketplace that connects freelancers with businesses or individuals needing specific services. The relevance of the gig economy is underscored by its rapid growth, fueled by technological advancements, changing workforce attitudes, and the increasing demand for flexibility in employment. It has redefined the way work is structured, executed, and perceived, impacting various sectors, from transportation and delivery services to professional and creative industries.

Parallel to the rise of the gig economy, there has been a significant evolution in the field of artificial intelligence (AI), particularly in Large Language Models (LLM), General AI, and Machine Learning (ML). These technologies are at the forefront of the current AI revolution, driving innovations across industries and altering the landscape of work and productivity.

Today’s article aims to delve into the intersection of the gig economy with the burgeoning fields of LLM, General AI, and ML. The exploration will be twofold. First, we will examine the synergies that these technological advancements create within the gig economy, including how AI and ML empower gig workers, enhance their efficiency, and open new opportunities. On the flip side, we will confront the potential conflicts and challenges that arise, such as potential job displacement, ethical concerns, and the further widening of the digital divide. 

The Gig Economy: An Overview

People often ask what I mean by the term “Gig Economy.” If I were to provide a definition, it would be as follows: an economic system where temporary, flexible jobs are commonplace, and companies tend to hire independent contractors and freelancers instead of full-time employees. 

There are 4 main types of gig work:

1. Digital and Creative Services: Platforms like Fiverr and Upwork have revolutionized the gig economy by facilitating connections between freelancers and clients globally. These platforms host a range of services, from graphic design to digital marketing and software development, allowing skilled professionals to offer their expertise on a project basis.

2. Professional Consulting and Expertise: Catalant is an example of a platform that caters to high-end professional consulting gigs. It connects businesses with experts in various fields for project-based work, ranging from market research to strategic planning.

3. Physical Task-Based Work: Beyond the digital realm, the gig economy also thrives in more traditional, physical spaces. Rideshare services like Uber and Lyft, food delivery services like DoorDash and Grubhub, and home service platforms like TaskRabbit are prime examples. These services allow individuals to work flexibly, providing transportation, food delivery, and home repair services on a per-task basis.

4. Specialized Skills and Trades: Much like the previous type of gig work, there are several specialized skills and trades, such as electricians, plumbers, and carpenters who can offer their services independently or through local service platforms.

Note: Physical-based work and specialized trades are not yet threatened by Gen-AI –although automation is on the horizon– so for the purposes of our analysis, we’ll only focus on gig work that is offered online. 

While the roots of the gig economy can be traced back to independent craftsmanship and freelance work, with the advent of the internet and mobile technology, it has expanded immensely. Particularly during the last decade, this sector has seen exponential growth driven by technological advancements and changing attitudes toward work-life balance. Current trends indicate a continued rise, with more professionals seeking career flexibility and autonomy. This shift is also fueled by economic factors, with companies increasingly relying on gig workers as a way to reduce operational costs and increase efficiency. A recent report by the World Bank outlines several of these trends: 

Growth of Online Gig Work: Demand for gig work increased by 41% between 2016 and the first quarter of 2023. 

This growth reflects the expansion of global online gig platforms, which tripled in number between 2010 and 2020. The COVID-19 pandemic further accelerated the use of digital platforms, showcasing increased demand for online labor after the initial drop in 2020 due to early pandemic impacts.

Seasonal Fluctuations in Demand: Despite the overall increase, demand for gig workers experiences seasonal fluctuations, peaking at the beginning and end of each year, with lower demand in the second and third quarters. This trend persisted across various occupations and regions, except for 2022, which may reflect the impact of the war in Ukraine.

Geographical Distribution of Demand: High-income countries (HICs) dominate the demand for online labor, contributing about 78% of global demand, with the United States alone accounting for nearly 4 in 10 vacancies. Lower-middle-income countries (LMICs) are the second most significant contributors, collectively accounting for 15.4% of global demand, with notable participation from India, Pakistan, the Philippines, Nigeria, and Ukraine.

Shifts in Global Demand: The share of global labor demand from developing countries like India and Pakistan has increased, indicating a growing contribution from these regions to the gig economy. Conversely, the share of global demand from the United States decreased by 10%, highlighting a shift toward more diversified global demand for gig work.

Employment Patterns in Gig Work: Online gig work is an important supplemental income source, with about 40% of gig workers engaging in it as a secondary activity. The intensity of gig work varies across regions, with a significant portion in East Asia and the Pacific treating it as their main occupation, while in South Asia, most gig workers engage in such work marginally.

Earnings and Income: On average, online gig workers in Bangladesh report earning significantly more from freelancing platforms compared to the average monthly household income. Similarly, in Pakistan, online gig workers earn substantially more than informal workers, indicating the potential of gig work to improve income levels.

But these upward trends are competing with similar trends on the technological front.

 

Technological Advances: LLM, General AI, and ML

The journey of AI and ML began in the mid-20th century, with foundational concepts like Alan Turing’s computational theory and the development of simple neural networks. The field experienced periods of ‘AI Winter’ where progress stalled, mainly due to limited computing power and data. However, the resurgence in the 21st century, fueled by advances in computational capacity and data availability, led to significant breakthroughs.

The last decade has witnessed remarkable advancements, particularly in deep learning, a subset of ML, which includes the development of sophisticated neural networks, leading to the creation of powerful LLMs and the gradual movement toward the conceptualization of General AI.

LLMs, such as OpenAI’s GPT series, are AI systems designed to understand, interpret, and generate human-like text. General AI refers to an advanced form of AI that can understand, learn, and apply its intelligence to a wide range of problems, much like human cognitive abilities. Although fully functional, General AI is not yet a reality, but its conceptual framework guides much of the current AI research.

These models and tools have multiple applications and parallel to the apprehensions surrounding them, it’s important to explore their potential and the promises they hold. 

LLM and General AI can empower gig workers by enhancing their efficiency and productivity in multiple ways:

LLM and ML algorithms have been instrumental in automating routine and administrative tasks. For example, chatbots powered by LLM can handle customer inquiries, scheduling, and basic administrative tasks, allowing gig workers to focus on more complex and revenue-generating activities.

ML algorithms provide gig workers with better tools for data-driven decision-making, allowing them to derive insights from large datasets. These insights can include optimal pricing strategies, peak demand periods, and customer preferences, enabling gig workers to make informed decisions that boost their productivity and income.

Furthermore, advanced AI tools aid in personalizing services and improving the quality of deliverables. For instance, freelance graphic designers use AI-powered design tools that suggest improvements, offer templates, and automate parts of the design process, enhancing both efficiency and output quality.

They help them improve skill development by providing Real-Time Feedback and Improvement. For creative professionals, AI tools provide real-time feedback on their work, suggesting improvements and offering tutorials for skill enhancement.

They can also be used to improve Language and Communication Tools: LLM-based language tools assist gig workers in overcoming language barriers, enabling them to communicate effectively with a diverse client base and expand their reach.

For example, freelance writers and content creators can use these tools to become more efficient. Utilizing LLMs like GPT-3 for research assistance, idea generation, and initial drafting, freelance writers have reported a significant increase in productivity, with some noting a 50% reduction in the time taken to complete articles.

Freelance graphic designers use ML-powered image editing tools that can automatically enhance photos, create complex image compositions, or even generate art. These tools significantly reduce the time and effort required for tasks like object removal, color correction, and layout design.

Freelancers in digital marketing utilize ML tools to analyze trends, engagement, and consumer behavior on social media platforms, enabling them to craft more effective marketing strategies. Freelance developers use AI-powered coding assistants that suggest optimizations, identify bugs, and offer code refactoring suggestions, making their coding process faster and more efficient. A freelance software developer used an AI-based code assistant to identify and fix bugs in a complex codebase, reducing the debugging time by 50%.

In each of these cases, gig workers are not just passive participants in a tech-driven economy; they are actively leveraging ML and AI to enhance their service quality, efficiency, and, ultimately, their competitiveness in the market.

Looking ahead, it’s safe to say that AI can revolutionize all human endeavors, albeit with some ethical and technical hurdles to overcome.

Conflicts and Challenges

Studies indicate that automation and AI can potentially displace a significant number of jobs, so the fear around these tools is merited. For the gig economy, the threat is twofold: automation of tasks (like data entry or basic customer service) and the development of sophisticated AI that can perform complex tasks (like writing or design work).

My colleague Daniel Rock and his co-authors try to add rigor to this discussion through their study, which explores the impact of advanced large language models like ChatGPT on the U.S. job market, particularly emphasizing the enhanced performance from LLM-integrated software. 

They first introduce a framework to evaluate job roles through their compatibility with LLM capabilities, incorporating expert human judgment alongside GPT-4 categorizations. 

This approach aims to understand how LLM advancements could reshape employment dynamics by aligning jobs with the technological strengths of LLMs and GPT-4:

“Our findings reveal that around 80% of the U.S. workforce could have at least 10% of their work tasks affected by the introduction of LLMs, while approximately 19% of workers may see at least 50% of their tasks impacted. We do not make predictions about the development or adoption timeline of such LLMs…. Our analysis suggests that, with access to an LLM, about 15% of all worker tasks in the US could be completed significantly faster at the same level of quality. When incorporating software and tooling built on top of LLMs, this share increases to between 47 and 56% of all tasks.”

The researchers suggest that software enhanced by LLMs will significantly amplify the economic impacts of these models. They infer that LLMs exhibit characteristics akin to general-purpose technologies, hinting at their potential to drive major economic, social, and policy changes.

The paper is not dedicated to (or even mentions) the gig economy, but let’s look at the jobs that are likely to be replaced by Gen-AI in terms of exposure:

Survey Researchers: 84.4%

Writers and Authors: 82.5%

Interpreters and Translators: 82.4%

Public Relations Specialists: 80.6%

Tax Preparers: 100%

Financial Quantitative Analysts: 100%

Web and Digital Interface Designers: 100%

Proofreaders and Copy Markers: 90.9%

Many are tasks currently offered by gig workers. A recent paper, “ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks,” uses 2,382 tweets to demonstrate that ChatGPT surpasses crowd-workers in several annotation tasks, such as detecting relevance, stance, topics, and frames. It was specifically noted that ChatGPT’s zero-shot accuracy outperforms that of crowd-workers in four out of five tasks, and ChatGPT’s intercoder agreement is higher than that of both crowd-workers and trained annotators across all tasks. Furthermore, it is worth noting the cost per annotation with ChatGPT is under $0.003, making it approximately twenty times more cost-effective than using MTurk. 

So, while there is a steady increase in demand for online gig work, in the foreseeable future, many of these tasks will be executed by ChatGPT models —some of these tasks are already being performed by ChatGPT, in a cheaper and better way. 

Ethical Considerations and Privacy Concerns

While this article focuses on the impact of Gen-AI on job replacement, there are other issues that shouldn’t be dismissed.

Research in AI ethics has raised concerns about algorithmic bias, where AI systems may perpetuate and amplify societal biases. Studies reveal instances of racial and gender biases in commercial AI systems, which could manifest as biased hiring algorithms or unfair job recommendations in the gig economy.

The use of AI and ML in gig platforms involves massive data collection, which raises privacy concerns. The European Union’s General Data Protection Regulation (GDPR) addresses some of these concerns, but challenges remain.

This is not about ChatGPT and the Gen-AI model, but rather the use of AI in general. And these concerns are real both in the online gig work platforms as well as the other types mentioned above. 

Future Outlook

There’s no question that AI and ML will lead to more efficient matching of gig workers with jobs, tailoring opportunities to individual skills and preferences, which in turn can lead to higher job satisfaction and better job outcomes. Personalized AI assistants could become a common tool for gig workers, helping with tasks like schedule management, client communication, and job searching. Unfortunately, these are the same jobs with the highest risks of being replaced by AI and Gen-AI tools, so while integrating Gen-AI models and ML in the gig economy presents opportunities, it also brings significant challenges.

Nevertheless, as the gig economy is predicted to continue its expansion, diversifying into new sectors and industries, advanced AI and ML technologies will likely enable the creation of new types of gig work, particularly in areas like AI training, virtual assistance, and personalized services.

The future will likely see a hybrid model with human creativity and AI efficiency working in tandem. Training programs and educational curricula will need to focus on skills that complement AI, such as creative problem-solving, emotional intelligence, and critical thinking.

But just like everything, continuous learning and adaptability will be key. Workers will need to stay abreast of technological advancements and continually update their skills. Policy initiatives could include ‘lifelong learning accounts,’ or similar concepts, encouraging ongoing skill development.

In conclusion, the future of the gig economy in the era of advanced AI and ML is poised to be dynamic. While it presents challenges, it also offers immense opportunities for innovation, personalization, and growth.

However, in the competition between the increased demand for such services and the evolution of Gen-AI, the Fiverr stock price seems to indicate that investors don’t believe there’s a bright future for such a platform, and if there is one, it’s too far in the future, and not worth betting on.

But I wouldn’t be so pessimistic.

Karl Popper, a philosopher of science, notably critiqued the notion of accurately predicting human growth and societal development in his work The Poverty of Historicism. He argued against historicism—the belief that history unfolds according to deterministic laws—highlighting instead the unpredictability of human history due to the evolving nature of human knowledge. Popper posited that since the future growth of scientific knowledge, which influences societal and technological advancements, cannot be predicted with precision, it renders the prediction of human history’s long-term course equally uncertain. He emphasized the significance of individual human agency and the unforeseen innovations that shape history.

History has shown us over again that we shouldn’t bet against people’s ability to innovate. After all, betting against human innovation is akin to expecting a Starbucks barista to spell your name correctly on the first try—endearingly hopeful but statistically improbable.