🫣AI Reality Check: Separating Hype from Reality
As someone who's passionate about Applied AI, I follow the latest developments in the field every single day.
📰A recent WSJ article points out some realities about the current state of generative AI innovation. (article link in comments)
The pace of improvement in generative AI (GenAI) is slowing, and its usefulness is limited by the high cost of running it. With companies having already trained their models on most of the internet, finding new data to improve them is becoming increasingly difficult. And, even the best proprietary AI models are converging on similar scores, making generative AI a commodity.
The article highlights three key challenges:
1️⃣ Data limitations: We're running out of data to train AI models, and synthetic data won't fill the gap.
2️⃣ Costly to run: AI models are expensive to run, eating into company margins and making it difficult for startups to sustain.
3️⃣ Limited headcount impact: As I constantly talk about, AI is a task killer not a job killer. Thus, it's unlikely to reduce payroll costs to offset the costs of adding this tech to their environment.
(Companies are scapegoating AI with layoff announcements. I will write about this soon.)
🚨Business Value Creation
The answer lies in hardcore focus on creating value with AI!
One powerful path is to build secondary models that amplify the impact of AI, making it more accessible and affordable for businesses and governments.
A secondary model makes AI more useful and actionable for a specific industry or application.
It’s like an expert in your field, who has spent years learning and understanding the intricacies of your business or mission.
The expert provides valuable insights and guidance, helps you to make better decisions, optimize your operations, improve customer experience, drive revenue, and more.
🛠️Applied AI Examples
🏦In financial institutions, secondary models can capture regulatory requirements, risk management policies, and customer data to facilitate the development of Al-powered solutions for fraud detection, credit scoring, and customer service.
🏥 In healthcare, secondary models can integrate electronic health records, patient data, and clinical guidelines to support the development of Al-powered solutions for disease diagnosis, treatment planning, and patient outcomes tracking.
By using secondary model architectures, we can unlock the potential of AI, with genuinely useful AI solutions.
⚠️If you’re building software, don’t build general use capabilities! It’s a waste of your time and money. Big tech already owns that space and is developing so fast that you’ll be caught up to and crushed. The riches are in the niches!
🔔Bottom line, let’s build practical AI solutions that drive real value, rather than chasing the hype.
#AppliedAI #SecondaryModels #PracticalAISolutions #innovation
