The State of Generative AI in 2024: Part One

This is part one of a two-part series on the good, bad and the ugly of Generative AI. First, we’ll talk of the great things that are being achieved in the field and how industries can benefit from Generative AI.

A short 200 years after Charles Babbage invented the first programmable computer, OpenAI released ChatGPT, a chatbot built on their GPT 3.5 foundational model. This was the inflection point of the decades long AI race to build machines that could do anything humans could do. Up until now we’ve only seen Artificial General Intelligence as science fiction, such as Arthur C. Clarke’s murderous computer HAL or George Lucas’ chirpy robot R2-D2.

Now we seem to be closer to our lofty goals than ever with Generative AI, with startups raising billions of dollars in funding to develop faster and more capable large language models. New companies are propping up each day, hyperscalers are poaching senior executives from startups (looking at you, Microsoft) and almost every consumer product now has “AI features”. From what I know of the dot-com bubble in the late 90s, I’d say this looks quite similar. So, are we in a bubble? We’ll only find out once it pops.

Last week was a busy time in the generative AI space with the launch of Meta AI, the company’s answer to Google’s Gemini and OpenAI’s ChatGPT. It is built on the initial iterations of its Llama 3 foundational model in 8B and 70B parameters, with a 400B size reportedly under development. Meta already claims that these models outperform other comparable LLMs in the market such as Google’s Gemma 7B, Mistral’s Mistral 7B and Anthropic’s Claude 3 Sonnet.

Beyond the mainstream consumer product launches, what’s really happening behind the scenes? Are industry leaders and large corporations really utilizing generative AI technologies within their businesses? Are there real business use cases for this technology beyond just chatbots?

Consulting firm McKinsey says yes. Generative AI’s impact on productivity could add the equivalent of $2.6 trillion to $4.4 trillion annually in value to the global economy, according to the company’s research estimates. This would increase the impact of all artificial intelligence by 15 to 40 percent, and about 75 percent of the value that generative AI use cases could deliver falls across four areas: Customer operations, marketing and sales, software engineering, and R&D.

Industry players are increasing their budgets to build expert generative AI teams, launch products for both internal and external use, in hopes of optimizing their businesses. According to an EY survey, more than 80% of retail and commercial banks have mobilized Generative AI teams, spending between $1-50 Mn in the training and development for these teams. Similarly, another survey by EY found that more than 70% of Indian retailer respondents plan to implement Generative AI into their operations within the next one year by partnering with external technology providers. Healthcare and life sciences, another industry in which generative AI is slated to have a large impact, is also on the forefront of adoption, with more than 75% respondents already using or planning to implement generative AI within a year.

With the growing interest if the industry in generative AI, few popular use cases include:

1.       Content Generation: By far the most popular use case for generative tools, LLMs can take in prompts and create synthetic content in the form of text, image and video.

2.       Customer Service Automation: The use of generative AI for use cases such as conversational AI and chatbots can automate and increase the efficiency of human customer service representatives. 

3.       Coding Assistant: One of the more dystopian use cases of the technology, generative AI has started to write reasonably simple lines of code, sparking fears of job losses for software engineers.

4.       Supply Chain Optimization: Using GenAI improves predictions for sourcing and procurement, logistics, transportation, and collaboration with suppliers. Companies can also predict future demand for specific products based on historical data, trends, and external factors.

5.       Fraud Detection: Generative AI excels in real-time inspection of vast amounts of datasets and can flag or block suspicious financial activities as they happen, thereby preventing potential economic loss. GenAI can create synthetic datasets using Data Augmentation, which adds robustness to the model, enabling it to detect not only patterns but also similar attacks that could be missed using traditional methods.

Now that we know that generative AI is here to stay and is actively being used by businesses around the world, in part two we’ll talk about the challenges, drawbacks and risks this technology poses.

Well, that’s all for now. It has been a long time since I’ve written something, please forgive any errors.