Sometimes training the model is difficult. We need to have at least a few different components to evaluate and understand the behavior of different users to have a very, very high accuracy in the models and to have a very good response to the tool. Therefore, the learning of the model is sometimes complex. We may need a different code validation tool to understand the behavior of the models.
Director of Business Development at a educational organization with 1,001-5,000 employees
Real User
Jan 3, 2024
In future releases, I would like to see a more flexible environment. It's a good product for customization and developing products. But when we need the most control over the delivery, Watson isn't the best. We can't fix everything because we're working with a machine that's creating a product. And the ability to go in-depth and tweak our model easily would be really nice.
I had not considered how the solution could be improved because I was focused on how it was helping me to solve my issues. If I consider how we want to use it in our organization, certain areas of improvement can be addressed. For instance, we want to use it with Generative AI, not like ChatGPT, but in a way intended for industrial use. It would be beneficial to incorporate more AI into the solution.
The supporting language is limited, and other languages could be added. In addition, because it is on cloud, we cannot have any sensitive information because of regulations by the central banks and the government cybersecurity-related issues. We do not use any transactional details, just manually driven ones.
Research Director, Network Security at a tech services company with 10,001+ employees
Real User
Jan 31, 2021
I haven't dealt with the solution as significantly in the last probably two or three years. That said, my last deeper dive into that was around the need for the product within the organization. I'm sure it's gotten better and better as the program has gotten better, however, early on, they relied heavily on building out these massive reference tables. That was a ton of the work that had to be done Honestly, I haven't seen any comparative report that has run the same data through two different artificial intelligence or machine learning capabilities to get something out of it. I would love to see that. The top three industries should go head to head to feed the same data and then get evaluated for accuracy. Something like that would be tremendously interesting. I don't know if something like that exists.
Data Science Lead at a mining and metals company with 10,001+ employees
Real User
Top 20
Nov 5, 2020
The solution needs to improve on its consumerization. They need to expand on it. Right now, they don't make it very easy. Scaling is limited in some use cases. They need to make it easier to expand in all aspects. Their service is a little bad sometimes.
AI Development Platforms are crucial tools enabling developers to build AI models with efficiency, flexibility, and scalability. They support machine learning, deep learning, and natural language processing, offering essential functionalities for smart application creation.These platforms provide a robust environment where developers can experiment with model training and testing, leveraging integrated tools and frameworks. They support a wide array of programming languages and offer features...
Sometimes training the model is difficult. We need to have at least a few different components to evaluate and understand the behavior of different users to have a very, very high accuracy in the models and to have a very good response to the tool. Therefore, the learning of the model is sometimes complex. We may need a different code validation tool to understand the behavior of the models.
In future releases, I would like to see a more flexible environment. It's a good product for customization and developing products. But when we need the most control over the delivery, Watson isn't the best. We can't fix everything because we're working with a machine that's creating a product. And the ability to go in-depth and tweak our model easily would be really nice.
I had not considered how the solution could be improved because I was focused on how it was helping me to solve my issues. If I consider how we want to use it in our organization, certain areas of improvement can be addressed. For instance, we want to use it with Generative AI, not like ChatGPT, but in a way intended for industrial use. It would be beneficial to incorporate more AI into the solution.
The supporting language is limited, and other languages could be added. In addition, because it is on cloud, we cannot have any sensitive information because of regulations by the central banks and the government cybersecurity-related issues. We do not use any transactional details, just manually driven ones.
They should add more GPU processing power to improve performance, especially when dealing with large amounts of data.
I haven't dealt with the solution as significantly in the last probably two or three years. That said, my last deeper dive into that was around the need for the product within the organization. I'm sure it's gotten better and better as the program has gotten better, however, early on, they relied heavily on building out these massive reference tables. That was a ton of the work that had to be done Honestly, I haven't seen any comparative report that has run the same data through two different artificial intelligence or machine learning capabilities to get something out of it. I would love to see that. The top three industries should go head to head to feed the same data and then get evaluated for accuracy. Something like that would be tremendously interesting. I don't know if something like that exists.
The solution needs to improve on its consumerization. They need to expand on it. Right now, they don't make it very easy. Scaling is limited in some use cases. They need to make it easier to expand in all aspects. Their service is a little bad sometimes.