Meamwhile I think the law should garantee a customer can speak to someone at a company who has enough power to take action on the entire customer account.
At least in France we passed a law that says that if you subscribed to a service online, you can unsubscribed online too (no idea if we're first, or last).
I am also annoyed by these online services with basically no customer service.
But, what do you mean by “take action on the entire customer account?” “Trick or bribe a human” is not a very difficult task, so it must be the case that there are some things a customer service agent can’t do.
Granting access, clearly a no-go for any service that might hold important private documents. But also, canceling an account on that sort of service could be pretty catastrophic… which means they can’t resolve problems in a direction that stops them from billing you…
California adopted the rule that unsubscribe has to be the same process as subscribe a few years ago, then recently it was adopted nationally.
For a while some sites only let you unsubscribe online if your address was in California, so people would change their address and then unsubscribe online.
"Amazon Q offers 40+ built-in connectors to popular enterprise applications and document repositories, including S3, Salesforce, Google Drive, Microsoft 365, ServiceNow, Gmail, Slack, Atlassian, and Zendesk"
Having a direct link to S3 + existing connectors built in feels like a strong competitive moat. Interested to see how far they can expand on this
> Having a direct link to S3 + existing connectors built in feels like a strong competitive moat.
Not really; the hard part is processing the data from each of these sources, not downloading the data from the source. Sure, any company that wants to compete is going to need to raise at least $30m just to get to the same minimal baseline level of support, but that was already the case even without this announcement.
In terms of whether this product will win the market, I honestly doubt it. Figuring out how to best process the data from each source to yield good results is going to be highly subjective, and Amazon's culture makes it unlikely that it will succeed here.
(As someone with a company that offers an API to pre-process email for LLM ingestion, fwiw.)
Fair. I'm biased as someone who has worked with public and private enterprise clients in F500 security/healthcare/government sectors where there's been reluctance to adopt some of these solutions. Amazon offering this natively might mean:
1) I won't have to rely on a smaller third party to have access to my data for connecting & processing that may or may not match my data regulatory requirements (compliance, protection, residency, etc) and
2) it's one less vendor to deal with at a price that's hard to beat in a time when many companies are consolidating systems spend.
It's probably because Amazon has all the money in the world they need to develop their products, but I would expect to launch a new product with some minimal but strong features to attract customers. Q offers 40+ built-in connectors from day zero. Like I can imagine the engineers/managers working on connector number 25: "Man, we have implemented already 24 connectors and we don't even know if the product will be a success or not...".
Weird that I read a story about OpenAI's mysterious "Q" model, open up Hacker News, and read about Amazon's Q model. It will be interesting to see the business expertise functionality that they tout.
I see a lot of these data providers investing in chat style interfaces too, the main plus with aws is all the extra data a Confluence chat interface won’t have for example. Not sure how you reconcile inconsistent data from say slack and confluence. If they get it right though, this will be the top of the stack for AI for a lot of companies
Interesting approach to cast the net so wide in a single release.
I mean, I can imagine it makes a lot of sense for a company to just dump a bunch of documents into S3 and then expect an LLM to be able to answer questions on that corpus. In some sense, you don't even really care about what is happening in the background, i.e. is it RAG, fine-tuning, LoRA, etc.
Also, I can imagine a debugging scenario for AWS where you might want an AI assistant to have access to your Cloudwatch, ECS, EC2, etc. so you can ask questions like "X service is down, what interesting logs/metrics are worth looking at more closely". And instead of the truly terrible AI "smart" alerting solutions you can play a game of 20 questions with a GPT-3.5 level LLM.
These services are the tip of the iceberg compared to what will come in the next couple of years. I bet Azure will have similar offerings very soon. Maybe Amazon is working here to beat them to the punch?
Readit News
At least in France we passed a law that says that if you subscribed to a service online, you can unsubscribed online too (no idea if we're first, or last).
But, what do you mean by “take action on the entire customer account?” “Trick or bribe a human” is not a very difficult task, so it must be the case that there are some things a customer service agent can’t do.
Granting access, clearly a no-go for any service that might hold important private documents. But also, canceling an account on that sort of service could be pretty catastrophic… which means they can’t resolve problems in a direction that stops them from billing you…
For a while some sites only let you unsubscribe online if your address was in California, so people would change their address and then unsubscribe online.
Having a direct link to S3 + existing connectors built in feels like a strong competitive moat. Interested to see how far they can expand on this
Not really; the hard part is processing the data from each of these sources, not downloading the data from the source. Sure, any company that wants to compete is going to need to raise at least $30m just to get to the same minimal baseline level of support, but that was already the case even without this announcement.
In terms of whether this product will win the market, I honestly doubt it. Figuring out how to best process the data from each source to yield good results is going to be highly subjective, and Amazon's culture makes it unlikely that it will succeed here.
(As someone with a company that offers an API to pre-process email for LLM ingestion, fwiw.)
It's probably because Amazon has all the money in the world they need to develop their products, but I would expect to launch a new product with some minimal but strong features to attract customers. Q offers 40+ built-in connectors from day zero. Like I can imagine the engineers/managers working on connector number 25: "Man, we have implemented already 24 connectors and we don't even know if the product will be a success or not...".
I mean, I can imagine it makes a lot of sense for a company to just dump a bunch of documents into S3 and then expect an LLM to be able to answer questions on that corpus. In some sense, you don't even really care about what is happening in the background, i.e. is it RAG, fine-tuning, LoRA, etc.
Also, I can imagine a debugging scenario for AWS where you might want an AI assistant to have access to your Cloudwatch, ECS, EC2, etc. so you can ask questions like "X service is down, what interesting logs/metrics are worth looking at more closely". And instead of the truly terrible AI "smart" alerting solutions you can play a game of 20 questions with a GPT-3.5 level LLM.
These services are the tip of the iceberg compared to what will come in the next couple of years. I bet Azure will have similar offerings very soon. Maybe Amazon is working here to beat them to the punch?
https://www.businessinsider.com/amazon-vp-ai-race-developers...