A company is examining the barriers to adopting generative AI. Which is the MOST appropriate reason that generative AI (especially foundation models from managed services) is considered relatively easy to start using?
Choosing the reason generative AI is easy to adopt (a low barrier to entry).
Because you can retrain from scratch on a small amount of your own data.
The benefit of a foundation model is precisely that training from scratch is unnecessary in the first place.
Retraining on a small amount of data is not the reason for the ease, and the explanation is incorrect.
Because GPU infrastructure is provided for free.
Even with a managed service, charges based on usage do apply.
The reason for the ease is that little upfront investment and no infrastructure build-out are needed, not that it is free, so it is incorrect.
Because you use a pretrained foundation model and can start with little upfront investment.
Correct. With a managed service, you can use a pretrained foundation model right away through an API, so you do not need to train a huge model yourself and can adopt it with little upfront investment. A low barrier to entry is a benefit of generative AI.
Because evaluating the model's accuracy becomes unnecessary.
Even when using a pretrained model, evaluation for your own use case is still necessary.
Skipping evaluation is not the reason for the ease; it rather creates quality risk, so it is incorrect.
The reason generative AI (managed services) has a low barrier to entry is that 'you can use a pretrained foundation model right away through an API, do not need to train a huge model yourself or provision GPUs, and can start with little upfront investment.' 'Training from scratch is required,' 'owning a GPU cluster is required,' and 'years of accumulated data are a prerequisite' all raise barriers and are the opposite of reasons for ease.