Analyzing Generative AI Costs
When implementing generative AI in business, it’s essential to understand the range of potential costs. An off-the-shelf generative AI solution could start at a few hundred dollars per month, while a bespoke model based on a fine-tuned open-source model may lead to a price tag of $190,000 or even higher.
The cost of generative AI is influenced by the specific tasks you want to enhance, the model that best suits those tasks, and the chosen implementation approach. To optimize expenses, you should carefully consider your project requirements, assess on-premises and cloud infrastructure expenses, and decide whether to hire in-house AI talent or outsource the project.
We have previously discussed the comparison of generative artificial intelligence (Gen AI) with traditional AI and its use cases across several industries. We have also examined the cost of building artificial intelligence systems and infrastructure and focused on machine learning (ML) costs.
Now, let’s delve into the cost of generative AI implementation in business.
Selecting the right model and implementation approach are key factors that can impact the cost of generative AI. When considering generative AI, it’s crucial to:
– Identify the business tasks to enhance with generative AI
– Choose a model suited for those tasks
– Assess available implementation approaches
Foundation models, large models trained on extensive data, underpin generative AI solutions, simplifying development and reducing costs. These models’ cognitive capabilities rely on the number of parameters, which are elements learned from training data. Other factors, such as training data quality, diversity, model architecture, and learning algorithms, also influence the models’ capabilities.
All generative AI models can be categorized into closed-source and open-source models. Closed-source models are developed by large tech companies, while open-source models have their source code, training techniques, and sometimes training data available for public use.
Upon considering generative AI implementation, businesses can opt for four primary approaches:
1. Using closed-source models without customization
2. Retraining commercially available solutions on corporate data
3. Using open-source models “as is”
4. Retraining open-source models on corporate data
Each approach comes with its own implications for generative AI pricing and customization.
The cost of commercially available generative AI tools is typically based on the number of characters or tokens in input or output text. Some solutions charge based on the number of characters, while others use token-based billing, depending on the model’s training and processing methods.
Overall, understanding the associated expenses and considering these factors will help businesses make informed decisions and navigate the rapidly evolving tech landscape.