AI Creator's Path News: Salesforce's Slack API change blocks LLM data access. What is the impact on data management? #SlackAPI #LLM #Salesforce
Video explanation
Is it true that Slack data can no longer be used for AI training? A clear explanation of Salesforce's policy change!
Hello, I'm John, who will be explaining the world of AI and technology next to you! Do you use "Slack" at work? It has become an indispensable tool for daily communication and information sharing in many companies. There has been a big change in Slack, so I will pick up the news and deliver it to you today. Salesforce, the company that operates Slack, announced that they will "partially restrict the use of data stored in Slack for learning AI (artificial intelligence)." Some of you may be worried, "What does that mean? Will my company's Slack data no longer be usable for AI?" Don't worry, this article should clear it up for you. Let's take a look together!
What's changed? Slack's new promises
To put it simply, this change is:It will become more difficult to send all the information (messages, files, etc.) stored in Slack to an AI, especially a smart AI called an LLM (large-scale language model), and have it learn it."about it.
A typical example of an LLM (large-scale language model) is ChatGPT, which we've been hearing about a lot lately. It's an AI that can converse naturally with humans, and can do a lot of things, like create sentences, answer questions, translate, and more. To make this LLM smarter, it needs to be fed a large amount of text data and have it "learn."
So, let's take a look at what has specifically changed in Slack.
- Changes to API Terms of Use: First of all, on May 2024, 5, Slack's terms of use for its "API" were updated. API is a bit technical, but it stands for "Application Programming Interface" and is like a "window" or "connection rules" for software or programs to exchange information. For example, it is used like a weather forecast app using the API of the Japan Meteorological Agency to get the latest weather information.
- Do not bulk export data: The new rules prohibit, in principle, using this API to extract large amounts of data from Slack at once (known as "bulk export" in technical terms). For example, downloading all the conversation logs from all channels for the past year at once.
- Prohibition on use for LLM studies: And here's the crux of the news: it's now explicitly prohibited to use data accessed via Slack's API to learn LLM. In other words, "Stop using Slack data to make LLM smarter."
Salesforce explains the change as "part of our efforts to improve data search and discovery experiences across the enterprise." Well, that may be a bit abstract and confusing.
So, what will happen with Slack's data usage in the future?
You may be worried, "Oh, so does that mean we can no longer use AI to analyze Slack data or put it to good use?" But don't worry. It seems the door is still open.
In conjunction with this change, Salesforce has announced:Real-Time Search API" is now offering a new API. This new API will allow you to search "within" the Slack platform. In other words, rather than handing over all of Slack's data to an external AI and having it "study" it, it seems to encourage users to "search" and find the information they need within the Slack box each time.
The advantage of this new API is that the processing is completed within Slack, which may increase security concerns. On the other hand, for companies that wanted to flexibly link with external high-performance AI, things may be a little different than before.
What kind of impact does it have? You might need to be a little careful.
This change in policy by Salesforce (Slack) may actually have a gradual impact on the way we work and how companies use AI.
Recently, many companies have started to view the huge amount of conversation data, shared files, and project progress information that accumulates daily on Slack as a "treasure trove," and there has been a growing movement to utilize this data. For example,
- An internal search system that uses AI to organize information scattered throughout the company, allowing users to quickly find the information they need.
- A tool that uses AI to automatically summarize discussions and decisions made on Slack and create meeting minutes
- An AI chatbot that uses AI to learn from past customer interactions and inquiry response history and automatically proposes answers to new inquiries.
As such, there are an increasing number of cases where companies develop their own AI systems or introduce AI apps provided by external companies (third parties) in order to improve the efficiency of their internal operations and create new value. There are likely many of these AIs that are becoming smarter by having LLM learn from Slack data.
With this change in the terms, such AI apps and systems may no longer be able to use Slack data in the same way as before. Developers may have to reconsider how they design their systems and how they handle data. Users may also find themselves wondering, "Oh, that convenient function I was using before is no longer available?"
Why did Salesforce make this change? (John's guess)
The official announcement states that this is to improve data search, but I, John, would like to dig a little deeper and speculate on some possible reasons (although this is my own opinion!).
- Enhanced data security and privacy protection: For companies, data on Slack often contains confidential and personal information. There may be a desire to prevent the risk of this important information being unintentionally passed on to external AI as learning data, leading to information leaks and privacy violations. Recently, there has been a growing voice around the world asking, "Is it okay to let AI learn everything?", so this move could be seen as a response to that trend.
- Guidance to in-house AI platform: Salesforce has its own powerful AI platform called "Einstein AI". It may be that they have a strategic intention to provide advanced AI functions that utilize Slack data through this Einstein AI, and to provide value to users within their own ecosystem. In other words, "Please use our (Salesforce's) AI for amazing AI functions that use Slack data!"
- Controlling the value of your data: The data stored in Slack is extremely valuable in itself, and by controlling how and to what extent this data is used within the company, it appears that the company is aiming to maintain and improve the value of the platform.
It's possible that these reasons are intertwined.
A word from John
Slack's change to its terms and conditions is not just a change in one company's policy; it seems to have highlighted a major issue: how companies should deal with and use their "company data" as AI technology rapidly evolves and becomes more deeply ingrained in our work and lives. Pursuing convenience, ensuring security, and promoting their own business strategies. Many companies are currently searching for a balance between these elements.
As users, it may be a good opportunity to keep an eye on the trends of these technology companies and take a moment to think about how our data is being handled in the tools and services we use every day. We welcome the convenience of the world, but I think it's very important to know what's going on behind the scenes!
This article is based on the following original articles and is summarized from the author's perspective:
Salesforce changes Slack API terms to block bulk data access
for LLMs