• AI Trailblazers
  • Posts
  • Navigating AI: Gen AI Imperfections, Stochastic Parrots, and more

Navigating AI: Gen AI Imperfections, Stochastic Parrots, and more

August 12, 2024

Welcome to this week’s AI Trailblazers Newsletter and thank you for making us your go-to source for the business of AI.

As summer heats up, so does the AI world. From stock market valuations to ROI discussions, the focus is on AI's viability and economic sense. When we asked ChatGPT if investing in AI is worth it, it replied, “Investing in AI can be worthwhile given its transformative potential, but it's crucial to stay informed, manage risks, and maintain a diversified investment strategy.” Sounds smart, and luckily, AI Trailblazers is here to bring you the latest insights. Discover more below.

Mark your calendars: the AI Trailblazers Growth Summit is happening on October 30th in New York. Purchase your tickets now at the early bird rate, and stay tuned for agenda details. In the meantime, check out highlights from our last summit, and email us if you're interested in partnering.

Gen AI Imperfections are Spoiling the Party

Despite the hype surrounding AI, significant doubts remain about its true impact and capabilities. Greg Rosalsky, examines how current AI models face ethical, reliability, and practical challenges that limit their effectiveness in replacing human jobs. Additionally, AI's development poses substantial environmental and economic concerns, raising questions about its long-term sustainability and benefits.

  • Questionable Intelligence and Ethical Concerns: Current AI models, including ChatGPT, may seem intelligent but are essentially sophisticated pattern recognizers, often copying copyrighted material without authorization. This has led to numerous high-profile lawsuits and concerns about systematic plagiarism. AI's reliance on human-generated data to function poses significant ethical and legal challenges.

  • AI Hallucinations and Reliability Issues: AI-generated falsehoods, or "hallucinations," are common, with studies showing error rates between 3% and 27%. These inaccuracies undermine AI's reliability, making it unsuitable for tasks requiring precision. The industry's efforts to reduce these errors are ongoing, but many experts believe the problem may never be fully resolved.

  • Limited Job Automation and Overstated Capabilities: Despite fears of job displacement, AI has struggled to replace humans in many roles, such as language translators and drive-through attendants. AI's practical applications remain limited, often resulting in subpar performance compared to human workers. Claims about AI's prowess, such as excelling in coding or taking bar exams, are often exaggerated and do not reflect consistent, real-world utility.

  • Environmental Impact and Economic Concerns: AI's development and operation require immense energy, raising environmental and economic sustainability issues. The need for specialized microchips and expansive data centers contributes to high costs and significant electricity consumption. Additionally, the anticipated economic benefits of AI, such as increased productivity, have yet to materialize, with many experts predicting only modest impacts on economic growth.

AI Trailblazer Takeaways: The AI revolution is still in the first inning but making unforced errors. The technology will get better with every new release and so will applications to aide productivity. It is worth noting however, that the human ability to adapt, think and construct is difficult to replicate even for the mightiest of apps.

Why Stochastic Parrots Can't Truly Think?

AI language models, often described as "stochastic parrots," face significant challenges in understanding and reasoning beyond mere pattern recognition. This Guardian article details how despite advancements, these models struggle with reversing factual relationships and handling complex, non-standard problems.

  • Stochastic Parrots and Lack of True Understanding: Linguist Emily Bender and computer scientist Timnit Gebru describe AI language models as "stochastic parrots," stitching together sequences of linguistic forms without understanding their meaning. Despite improvements with more data, AI like ChatGPT lacks true intelligence, reasoning, or thought. This fundamental limitation raises questions about the true capabilities of AI models.

  • Reversal Curse and Relationship Understanding: AI models struggle with the "Reversal Curse," failing to reverse simple factual relationships, such as identifying a person's parent from a given name. This issue arises because AI learns relationships between tokens, not facts. As a result, models like GPT-4 can answer direct questions but falter when the information is reversed or presented differently.

  • Human Recall vs. AI Reasoning: Unlike humans, who can symmetrically reason relationships between facts, AI models have asymmetric recall capabilities. Humans can easily reverse known relationships, but AI models cannot infer reversed facts if they haven't been explicitly trained on them. This limitation underscores the fundamental differences between human cognitive abilities and AI's pattern recognition.

  • Struggles with Outliers and Complex Problems: AI models, including the most advanced ones, struggle with outlier scenarios and non-standard problems. Examples include common puzzles and subtle variations of word problems that confuse AI systems. This inability to handle unusual circumstances suggests that current AI approaches are limited and may fail to meet the high expectations set by proponents of artificial general intelligence (AGI).

AI Trailblazer Takeaways: Today, AI is struggling to replicate the decision making and processing ability of the human mind. Not surprising given that the human brain is a very complex series of neural networks evolved over millions of years. AI will get better over time….but how will it deal with emotions? Oiy!

Artificial Movie Intelligence

The rise of AI in Hollywood has sparked significant debate and division within the industry. While unions push for strict regulations to protect artists' rights, studios explore AI's potential in content creation. The Hollywood Reporter delves into the contrasting views on AI legislation, ongoing legal and ethical debates, and the mixed reactions from filmmakers regarding AI's role in their craft.

  • AI's Impact on Hollywood: The introduction of AI in Hollywood, notably through OpenAI's ChatGPT in 2022, has significantly affected the industry. AI-generated content, such as the song "Heart on my Sleeve," brought immediate attention to the technology's potential and risks. This led to mobilization efforts from unions like SAG-AFTRA and the Recording Industry Association of America to establish legislative protections against unauthorized use of AI in creative works.

  • Diverging Views on AI Legislation: While unions advocate for stringent regulations to protect artists' rights, studios are more cautious, supporting measures that don't hinder their ability to use "digital replicas" in content creation. The Motion Picture Association (MPA) appreciates legislative safeguards that protect free speech. Studios are also concerned with both protecting their intellectual property from AI misuse and exploring AI's potential in future content production.

  • Ongoing Legal and Ethical Debates: The U.S. Copyright Office has become a battleground for discussions on AI's intersection with intellectual property. Unions like the Writers Guild of America (WGA) and Directors Guild of America (DGA) push for laws requiring consent for training AI on copyrighted material. Studios argue for looser standards to allow AI-generated works to receive copyright protection, highlighting a fundamental disagreement on the role and regulation of AI in creative industries.

  • Adoption and Resistance Among Filmmakers: AI tools, especially OpenAI's Sora, have divided filmmakers, with some embracing AI as a useful tool and others criticizing it for undermining the integrity of filmmaking. Unions stress that AI cannot create in the same way humans can and advocate for moral rights for creators. The DGA acknowledges AI's potential benefits but insists on proper guardrails to ensure it supports rather than disrupts the filmmaking process.

AI Trailblazer Takeaways: During the 1920’s there was a big uproar when sound was introduced to movies. Filmmakers and critics worried that dialog would undermine the aesthetic of silent films. Well today, who wants to watch a movie with no sound or dialog? AI will follow a similar path in its adoption within the movie industry no doubt.

What is AI Trailblazers?

AI Trailblazers is a vibrant platform dedicated to uniting corporate marketers, technologists, entrepreneurs, and venture capitalists at the forefront of artificial intelligence (AI). We want to thank everyone who joined us for the incredibly successful AI Trailblazers Summer Summit in June in New York. We’ve compiled a short video to give you a glimpse into the day for those that couldn’t join us. Our next forum will be on October 30th! Purchase your tickets now at the early bird rate.

Our mission is to fuel growth, innovation, and career development among our members, who all want to be at the forefront of incorporating artificial intelligence into their businesses and their lives for strategic advantage. More information here.

Partner with AI Trailblazers

Our community of business leaders is growing fast and we are planning more forums. To discuss partnership opportunities that connect our partners with AI decision makers including at our summits and this newsletter, email us. Our AI Trailblazers Growth Summit this October is already shaping up to be something extra special.

Quote of the Week

“We are in this phase where we have to deeply work and make sure on these use cases, on these workflows, we are driving deeper progress on unlocking value, which I'm very bullish will happen. But these things take time. But if I were to take a longer-term outlook, I definitely see a big opportunity here.”

- Alphabet CEO Sundar Pichai