Generative AI: The Future is Here and It's Writing Itself

Some insights on generative AI from OCRT by Anton G. .

As we step into the era of advanced Artificial Intelligence (AI), the release of OpenAI's Generative Pre-Trained Transformer 4 (GPT-4) has sparked excitement and anticipation in the field of generative AI. GPT-4 promises to be a significant improvement over its predecessor, ChatGPT-3.5, with the ability to comprehend and interpret images, making it a powerful tool for generating detailed and accurate responses. However, with the cost of entry into language model development already reducing, competition is becoming increasingly fierce among industry giants such as Meta and Google, who are also vying for dominance in the space. As companies seek to leverage generative AI technologies to gain a competitive edge, it is crucial to consider the ethical implications and risks of developing and deploying generative AI tools. Furthermore, the implementation of generative AI raises complex questions about the future of work and the balance between disruption and opportunities. In this article, we explore the challenges and opportunities presented by generative AI technologies and discuss how industry can find the balance between innovation and responsibility.

GPT-4

One of the most significant updates to GPT-4 is its ability to comprehend and interpret images, allowing it to accept text and image inputs. This represents a significant improvement over ChatGPT-3.5, which only accepted text inputs and a magnitude improvement since its launch in November 2022. Although GPT-4 is limited to responding via text, its ability to process visual inputs such as images, graphs, and screenshots presents as a valuable tool for generating detailed and humanlike responses.

GPT-4 has been shown to produce highly nuanced and detailed output, making it ideal for a wide range of applications such as essay writing, program coding, and financial analysis.[i] However, as the technology advances, there is growing concern that higher-paying jobs, which involve more software-based tasks, may be at greater risk of being disrupted by AI-powered chatbots.

Available through ChatGPT Plus, Microsoft has already announced that it has begun integrating GPT-4 experiences into Bing Chat,[ii] paving the way for greater accessibility in the future. Due to the introduction of an advanced memory-context window that allows GPT-4 to contextualise information, the new iteration has improved its ability to generate coherent and meaningful language output. GPT-4's memory-context window represents a significant enhancement in its natural language processing capabilities compared to previous model iterations, which relied solely on attention mechanisms to process information.

GPT-4's extended short-term memory, which can keep up with up to 64,000 words, compared to ChatGPT-3.5's 8,000-word memory, also marks a significant advancement in natural language processing capabilities. In contrast to ChatGPT-3.5's text-to-text approach, the model's data-to-text approach further enhances its capabilities in generating natural language text, particularly in complex tasks such as question-answering, summarisation, and dialogue generation.[iii]

Another significant advantage of GPT-4 is its multilingual capability, which outperforms the English-language performance of its predecessor and other Large Language Models (LLM) in 24 out of 26 tested languages. GPT-4 can accurately answer thousands of multiple-choice questions across 26 languages, including low-resource languages such as Latvian, Welsh, and Swahili, making it a valuable tool for global businesses and organisations.

Finally, OpenAI has taken significant steps to improve the safety properties of GPT-4, perceivably making it safer than its predecessor. The model-level interventions increase the difficulty of eliciting inappropriate behaviour, presenting GPT-4 as a safer option for users. Overall, like an iPhone advertisement, all these improvements sound great but require additional analysis to understand second the third order effects on industry.

The cost of entry into language model development is drastically reducing.

The newly established market position of OpenAI is already being increasingly contested by formidable rivals such as Meta (the parent company of Facebook) and Google, as they engage in a competitive struggle for pre-eminence within the language model sphere.

Meta's Language Learning and Modelling Assistant (LLaMA) is a recent addition to the field.[iv] LLaMA is designed to run on a single graphics processing unit (GPU) and has been shown to outperform ChatGPT-3 on various benchmarks. Similarly, Google recently released its own proprietary AI technology called Bard, a Language Model for Dialogue Application (LaMDA) that uses natural language processing and deep learning algorithms to enhance the AI's understanding and response to user input.[v]

While Meta's LLaMA and Google's Bard are both Large Language Models (LLMs) developed by different companies with different approaches to natural language processing, it is evident that the cost of entry for developing such models is drastically reducing. For instance, outside of Google's PaLM (Pathways Language Model) and DeepMind's Chinchilla AI,[vi] Stanford University researchers recently built their own language model called Alpaca for less than USD $600, using an open-source language model called LLaMA 7B, which they fine-tuned with the help of ChatGPT-3.5.[vii] Surprisingly, Alpaca has performed similarly to ChatGPT-3.5 on many tasks, emphasising the increasing ease and affordability of replicating LLMs.

The emergence of Alpaca and its affordability compared to ChatGPT highlights the rapid pace of innovation in the generative AI industry. As more researchers and companies enter the field, innovation will likely accelerate, leading to increasingly powerful generative AI models. This presents both opportunities and challenges, and it is crucial for policymakers to address these issues and ensure that ethical considerations and the public good guide AI development.

Finding the Competitive Edge: How Industry Can Make the Most of AI Technologies.

The advent of AI technologies such as deep learning and Generative Adversarial Networks (GAN) has brought about a transformational shift in how industries operate.[viii] Industry can now leverage generative AI capabilities at scale to reduce costs, save time, and gain a competitive edge. However, with such powerful tools accessible to competitors as well, stakeholders must strategically identify use cases that can create a true competitive advantage and significant impact relative to existing solutions and incurred adoption process costs.

The revolutionary potential of generative AI can be comprehensively explained through three key functional characteristics. Firstly, generative AI exhibits seemingly boundless memory and remarkable pattern recognition capabilities. Secondly, it demonstrates low-code or no-code properties, which revolutionise traditional programming. Finally, generative AI has the potential to augment many roles by increasing productivity, performance, and creativity.

With the potential presented by generative AI, organisations worldwide are already actively working towards enhancing the reliability of outputs by employing reinforcement learning with human feedback and exploring alternative approaches that combine generative AI with traditional AI and machine learning. Significant improvements in the performance of generative AI are expected to materialise soon, with some forecasts indicating that this technology will be capable of producing final-draft content by 2030.[ix]

Opportunities can be found throughout the value chain, from improving offerings and reducing time-to-market to cost savings and creating entirely new ideas. Organisations must make strategic decisions about whether to fine-tune existing LLMs or train a custom model to achieve success. Fine-tuning presents a more cost-effective option that can jumpstart experimentation, while training a custom LLM offers greater flexibility but comes with higher costs and capability requirements. Decision makers must carefully assess the timing of such an investment and weigh the potential costs of moving too soon on a complex project for which talent and technology are not yet ready against the risks of falling behind.

Additionally, protecting industry from the risks of generative AI requires policies that limit its use to well-established guardrails and ensure data ownership and review processes to prevent harmful content from being published. Organisations must also train employees on the safe and appropriate use of generative AI and caution against overconfidence in its capabilities. Decision-makers can adapt recommendations for responsible publication and implement mechanisms like licensing for downstream uses to manage the lack of a truth function in generative AI.

By crafting a proactive strategic approach to generative AI, industry can distinguish the market signals from the market noise and create long-term competitive advantages. However, this requires a nuanced understanding of the technology's capabilities and limitations and thoughtful planning and collaboration across entire enterprises.

Competition and Profit vs Safety: Risks in Developing AI Tools

The increasing availability of LLMs and the widespread adoption of generative AI technology in various industries, including customer service and search engines, have many implications for the future of work. However, concerns have been raised about the potential negative impacts of LLMs, such as spam, misinformation, malware creation, and targeted harassment. While current market participants are aware of these issues and work to curtail these capabilities, there are concerns that governments need to move more quickly to regulate the use of AI for the public good.

The USD $600 Alpaca development is a prime example of rapid innovation within the AI industry, particularly in LLMs. While these models can potentially revolutionise many aspects of society, they also pose significant risks that require attention from policymakers and industry leaders. The ethical implications of AI development and deployment are complex and multifaceted. It is critical to consider the nuances of the issues to ensure that the technology is leveraged to promote the greater good.

The potential for generative AI misuse is a concern, especially in the context of actors who may not prioritise safety measures. While the beneficial impacts of AI on society are undeniable, it is increasingly important to ensure that the technology is developed and deployed ethically and responsibly. As it stands, any regulation of AI is currently voluntary, underscoring the situation's urgency.[x] The prospect of large-scale disinformation campaigns and offensive cyberattacks is a pressing issue that demands attention to ensure the public's safety.

The intense competition among companies developing generative AI tools and the race for market penetration may incentivise some to prioritise profit over safety, which is concerning. The recent use of voice-cloning generative AI tools by phone scammers to defraud victims serves as an example of the potential harm that can result from the misuse of such technology.[xi] Therefore, it is crucial to establish industry-wide safety standards and regulations that prioritise public safety and hold companies accountable for the responsible deployment of AI technologies.

Regulation of generative AI is a complex and time-consuming process, and the industry's fast-paced growth may outpace standard regulatory frameworks for some time to come. However, it is imperative to take decisive action to prevent the technology from being misused for nefarious purposes. This will require a multifaceted approach involving industry leaders, government bodies, researchers, and stakeholders from various fields.

Complexities of AI in the Workforce: Balancing Disruption and Opportunities.

The impact of generative AI on the workforce has been a concern for many experts, as the technology can potentially replace certain jobs, particularly those in white-collar professions. Recent research has identified ten specific jobs that are most at risk of being replaced by generative AI, including tech jobs such as coders, software developers, and data analysts, media jobs such as advertising, technical writing, journalism, and content creation, legal industry jobs such as paralegals and legal assistants, customer service, and administrative roles.[xii] These jobs typically involve tasks that can be automated using generative AI, such as data analysis, content generation, and customer interaction.[xiii]

According to a recent US study, the implementation of generative AI can significantly impact the workforce, with up to 80% of workers potentially facing changes to some of their work tasks.[xiv] Around 19% of workers could experience a significant impact, with at least 50% of their tasks being affected.[xv] This is due to recent advancements in generative AI, natural language processing, and biometrics, which are accelerating efforts to automate certain job tasks, sometimes entirely. However, while generative AI has the potential to transform many industries, it is important to note that the technology is still in its early stages, and there are limitations to what it can do. While generative AI's most significant impact currently comes from changing jobs rather than replacing them, it is accessed that up to 25% of work activities in the US across all occupations could be automated by 2030.[xvi]

However, it is unlikely that generative AI will soon replace humans in the workplace, as it lacks the cognitive flexibility, creativity, and common-sense reasoning abilities that people possess. Instead, generative AI is more likely to be used as a productivity-enhancing tool, freeing employees to focus on more complex tasks requiring human skills, such as creativity and critical thinking. By automating repetitive tasks, generative AI could likely enhance certain jobs, such as coding and programming.

Notably, the accuracy of generative AI systems also significantly depends on the quality of the data and the algorithms used to process it. While generative AI can process and analyse vast amounts of data at a speed that people cannot match, it is not infallible. It can sometimes produce unintended errors or generate misinformation. Therefore, it is important to approach the integration of generative AI into certain industries with caution and to ensure that systems are in place to monitor and correct errors.

Preparing the workforce for the advent of generative AI requires a collaborative effort between leadership teams, human resources, and employees themselves. Redefining impacted roles and responsibilities will likely become essential to this assimilation. While traditional AI and machine-learning algorithms have already enabled people to work more autonomously, generative AI can further augment many roles by increasing productivity, performance, and creativity. However, these changes cannot occur in isolation.

Overall, while generative AI has the potential to disrupt certain industries and job roles, it is important to approach the topic with nuance and understand the limitations of the technology. When using generative AI, human judgment is still needed to avoid error and bias. Industry must remain mindful of its potential impact on the workforce as potential uses are explored.

Furthermore, integrating generative AI into certain industries could enhance job opportunities rather than replace them. Generative AI can identify new trends and opportunities, develop innovative solutions, and improve the quality of products and services. This could lead to creating new jobs and enhancing existing ones, as employees are empowered to work with generative AI technologies to achieve more efficient and effective outcomes.

Conclusion

The emergence of ChatGPT in late 2022 has sparked a surge in productivity hacks and early adoption, but the potential of generative AI goes far beyond immediate gains and technical limitations.[xvii] Today it has the power to revolutionise traditional business models and disrupt nearly every industry, offering both competitive advantage and creative destruction. As a result, organisational leaders must develop a comprehensive generative AI strategy that addresses the ethical implications of this technology, such as job displacement and potential misuse.

To effectively leverage the benefits of generative AI while prioritising employee well-being and collaboration, industry leaders must prioritise three critical aspects in their strategy: innovative possibilities, ethical considerations, and long-term impact. The first aspect is exploring the innovative possibilities that arise when every employee has access to the remarkable pattern recognition capabilities and seemingly infinite memory of generative AI. This raises questions about how this technology will change the way employees' roles are defined and managed and highlights the importance of human creativity in the face of the potential for AI-generated sameness.

The second aspect focuses on ethical considerations in developing and using generative AI. While this technology holds vast potential, it also raises important ethical questions regarding its impact on job displacement and the potential for misuse. Engaging in ongoing research and dialogue is crucial to ensure that the development and deployment of these language models are regulated, responsible, and ethical.

The third and final aspect considers the long-term impact of generative AI on business models. As generative AI becomes increasingly prevalent in the workplace, industry must develop a forward-thinking, adaptive, and ethical approach that provides a competitive advantage and creates new opportunities for growth. Industry leaders must ensure that humans review and edit AI-generated content to ensure accuracy and avoid copyright infringement. By prioritising employee well-being and collaboration while leveraging the potential benefits of generative AI, organisations can craft a strategic approach to generative AI that is both innovative and responsible.

[i] https://au.pcmag.com/news/99281/openai-chatgpt-could-disrupt-19-of-us-jobs-is-yours-on-the-list published 21 Mar 23

[ii] GPT-4 was just announced, and Microsoft confirmed that it powers the new Bing | Windows Central, published on 15 Mar 23

[iii] https://www.theatlantic.com/technology/archive/2023/03/gpt-4-has-memory-context-window/673426/, published 18?Mar 23

[iv] https://aibusiness.com/meta/meta-s-llama-language-model-outperforms-openai-s-gpt-3, published 28 Feb 23

[v] https://www.theguardian.com/technology/2023/mar/21/googles-bard-chatbot-launches-in-us-and-uk, published 22 Mar 23

[vi] What Is Chinchilla AI: Chatbot Language Model Rival By Deepmind To GPT-3 - Dataconomy, published 02 Feb 23

[vii] https://newatlas.com/technology/stanford-alpaca-cheap-gpt/, published 19 Mar 23

[viii] Guide to Generative Adversarial Networks (GANs) in 2023 - viso.ai, accessed 21 Mar 23

[ix] https://www.bcg.com/publications/2023/ceo-guide-to-ai-revolution, published 07 Mar 23

[x] https://fortune.com/2023/03/18/openai-ceo-sam-altman-warns-that-other-ai-developers-working-on-chatgpt-like-tools-wont-put-on-safety-limits-and-clock-is-ticking/, published 19 Mar 23

[xi] https://www.cbsnews.com/news/ai-scam-voice-cloning-rising/, published 21 Mar 23

[xii] https://www.businessinsider.com/chatgpt-jobs-at-risk-replacement-artificial-intelligence-ai-labor-trends-2023-02#tech-jobs-coders-computer-programmers-software-engineers-data-analysts-1, published 03 Feb 23

[xiii] https://www.washingtonpost.com/technology/interactive/2023/ai-jobs-workplace/, published 20 Amr 23

[xiv] GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models, University of Pennsylvania, published 21 Mar 23

[xv] https://www.vice.com/en/article/g5ypy4/openai-research-says-80-of-us-workers-will-have-jobs-impacted-by-gpt, published 21 Mar 23

[xvi] https://www.wsj.com/articles/ai-chatgpt-chatbot-workplace-call-centers-5cd2142a, published 18 Feb 23

[xvii] https://www.washingtonpost.com/business/2023/03/20/competing-with-ai-in-the-workplace-is-fueling-incivility-at-work/2bddcef8-c70f-11ed-9cc5-a58a4f6d84cd_story.html, published 20 Mar 23