.svg)
AI has been around for decades, but the recent surge in interest stands out. It’s impossible to escape the discourse around generative AI – and platforms like ChatGPT are setting records, outpacing the fastest-growing consumer technologies in history. There’s no denying these technologies are radically transforming our world, but are there no obstacles to their adoption? Will this growth curve continue, unabated?
Not so fast. The breathtaking evolution of generative AI doesn’t come without speed bumps. Trust, transparency, and cost pose considerable challenges to the success and scale of these innovations.
The Trust & Accuracy Dilemma
Large language model (LLM) chatbots like ChatGPT demonstrate incredible capabilities, but they’re far from perfect.
These platforms have a tendency to ‘hallucinate’, fabricating information and confidently presenting falsehoods as truths. Even Google’s Bard, trained with an extra-cautious lens and the ability to browse the internet, is not immune to such errors.
Reports like the New York Times piece about a GPT-authored legal brief containing false legal precedents further shake public trust. For widespread adoption, it’s crucial that these platforms be reliable sources of information – or at least learn to accurately acknowledge their limitations.
Opening the Black Box: Transparency, Control & Bias
LLMs are powerful tools, but their decision-making patterns remain opaque. When asked to explain or justify their responses, chatbots often falter, providing disjointed explanations or even fabricating citations. Until we understand these ‘black boxes’, many users may hesitate to rely on their logic.
Moreover, the data sets on which LLMs are trained raise questions about how inherent biases could influence platforms’ responses. Increased transparency is pivotal to building public trust and easing concerns about possible bias.
Data Privacy Fears Loom Large
Data privacy is another critical concern, particularly for businesses. After an incident at Samsung where employees inadvertently entered sensitive code into ChatGPT, the company barred its staff from using generative AI platforms to prevent potential data leaks. Many companies across the financial, tech, and legal sectors have implemented similar bans.
To win enterprise trust, platforms must assure users their data won’t be misused or shared without consent.
The High Cost of AI
One of the key challenges in scaling these platforms is the immense cost involved in training and operating an LLM. The latest models use more than 500 billion parameters, and future models are likely to run into the trillions. As OpenAI’s CEO, Sam Altman, noted in December 2022, each query can cost an LLM platform ‘single-digit cents.’
While tech giants like Google can bear these high costs, startups and smaller businesses may struggle to sustain their own platforms. Given the sustainability of their current margin structure, traditional search platforms – less than 100 times as costly as LLM chatbots – are unlikely to be replaced in the short term.
—
These hurdles are significant, but they shouldn’t blind us to AI’s potential. Public appetite for commercial AI platforms is at an all-time high, and as our AI infrastructure matures, the market for these products will surge into the trillions. The growth of these platforms may hinge on their ability to overcome these obstacles and more, but if they navigate the terrain carefully, generative AI could become the most impactful commercial technology in modern history.
