Generative AI Challenges and Limitations for AWS Customers

Customers at the AWS re:Invent conference last week explored the challenges of embedding GenAI into apps and processes in interviews and panel discussions.

[{"selector":"#anim-35b19f5d-ff33-40d0-a678-df002687bd2a","keyframes":{"transform":["translate3d(-121.79931%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-75b1e47b-f48b-4d42-ac42-fe8c8f0de0b2","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-c48ccb1a-ac7b-4e49-8333-802ab061b904","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

Concerns about the safety of the huge language models behind the AI capable of replicating human-generated material exacerbated the difficulty.

[{"selector":"#anim-3a58c51e-555d-4847-801a-51b62d615618","keyframes":{"transform":["translate3d(-124.72727%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-d8b4328b-e945-4fef-b93e-8b56eaed8de6","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-ca9add76-ab9c-4d2f-960a-f3a03e97db37","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

None of AWS's subscribers or partners were dissatisfied with the cloud provider's three-tier AI approach, which included infrastructure, platform and tools, and applications.

[{"selector":"#anim-bd17d5f2-7aed-4562-a95d-b37e681aaf08","keyframes":{"transform":["translate3d(-117.0418%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-ed57da12-9951-4886-b098-7b4dc77b2ad2","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-f0a02067-b9df-4a53-bd64-d14e190f606c","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

Many people considered its Bedrock platform to be a substantial development in selecting, training, and fine-tuning an LLM.

[{"selector":"#anim-26999f55-33eb-4603-8163-521078e5ecf7","keyframes":{"transform":["translate3d(-123.40426%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-4fa514ec-47c3-4e48-9dee-823b8ccf7685","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-53918177-b34a-42bf-8c54-6bbdf7ff1473","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

Nonetheless, organizations are concerned about failing to adapt to technical immaturity, manage data security and governance, avoid biases introduced into models, and correct erroneous LLM answers.

[{"selector":"#anim-712f6f80-7e64-442c-b9fb-df781704b2e1","keyframes":{"transform":["translate3d(-119.93243%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-a28ca982-f115-483d-baad-f307143ee90a","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-4aa4cfb9-75ec-49ee-a2a5-ec666dc7dfe5","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

Generative AI Challenges and Limitations for AWS Customers

Customers at the AWS re:Invent conference last week explored the challenges of embedding GenAI into apps and processes in interviews and panel discussions.

Concerns about the safety of the huge language models behind the AI capable of replicating human-generated material exacerbated the difficulty.

None of AWS's subscribers or partners were dissatisfied with the cloud provider's three-tier AI approach, which included infrastructure, platform and tools, and applications.

Many people considered its Bedrock platform to be a substantial development in selecting, training, and fine-tuning an LLM.

Nonetheless, organizations are concerned about failing to adapt to technical immaturity, manage data security and governance, avoid biases introduced into models, and correct erroneous LLM answers.

Read more stories.