But now it\u2019s not just humans who can take advantage of this technology; it\u2019s AIs\u2014and in particular LLMs\u2014as well. LLM foundation models are powerful. But LLM foundation models with our foundation tool are even more so. And with the maturing of LLMs we\u2019re finally now in a position to provide to LLMs access to Wolfram tech in a standard, general way.<\/p>\n
It is, I believe, an important moment of convergence. My concept over the decades has been to build very broad and general technology\u2014which is now a perfect fit for the breadth of LLM foundation models. LLMs can call specific specialized tools, and that will be useful for plenty of specific specialized purposes. But what Wolfram Language uniquely represents is a general tool\u2014with general access to the great power that precise computation and knowledge bring. <\/p>\n
But there\u2019s actually also much more. I designed Wolfram Language from the beginning to be a powerful medium not only for doing computation but also for representing and thinking about things computationally<\/a>. I\u2019d always assumed I was doing this for humans. But it now turns out that AIs need the same things\u2014and that Wolfram Language provides the perfect medium for AIs to \u201cthink\u201d and \u201creason\u201d computationally.<\/p>\n There\u2019s another point as well. In its effort to make as much as possible computable, Wolfram Language not only has an immense amount inside, but also provides a uniquely unified hub for connecting to other systems and services<\/a>. And that\u2019s part of why it\u2019s now possible to make such an effective connection between LLM foundation models and the foundation tool that is the Wolfram Language.<\/p>\n On January 9, 2023, just weeks after ChatGPT burst onto the scene, I posted a piece entitled \u201cWolfram|Alpha as the Way to Bring Computational Knowledge Superpowers to ChatGPT<\/a>\u201d. Two months later we released the first Wolfram plugin for ChatGPT<\/a> (and in between I wrote what quickly became a rather popular little book entitled What Is ChatGPT Doing \u2026 and Why Does It Work?<\/a><\/em>). The plugin was a modest but good start. But at the time LLMs and the ecosystem around them weren\u2019t really ready for the bigger story. <\/p>\n Would LLMs even in the end need tools at all? Or\u2014despite the fundamental issues that seemed at least to me scientifically rather clear right from the start\u2014would LLMs somehow magically find a way to do deep computation themselves? Or to guarantee to get precise, reliable results? And even if LLMs were going to use tools, how would that process be engineered, and what would the deployment model for it be?<\/p>\n Three years have now passed, and much has clarified. The core capabilities of LLMs have come into better focus (even though there\u2019s a lot we still don\u2019t know scientifically about them). And it\u2019s become much clearer that\u2014at least for the modalities LLMs currently address\u2014most of the growth in their practical value is going to have to do with how they are harnessed and connected. And this understanding highlights more than ever the broad importance of providing LLMs with the foundation tool that our technology represents.<\/p>\n And the good news is that there are now streamlined ways to do this\u2014using protocols and methods that have emerged around LLMs, and using new technology that we\u2019ve developed. The tighter the integration between foundation models and our foundation tool, the more powerful the combination will be. Ultimately it\u2019ll be a story of aligning the pre-training and core engineering of LLMs with our foundation tool. But an approach that\u2019s immediately and broadly applicable today\u2014and for which we\u2019re releasing several new products\u2014is based on what we call computation-augmented generation, or CAG. <\/p>\n The key idea of CAG is to inject in real time capabilities from our foundation tool into the stream of content that LLMs generate. In traditional retrieval-augmented generation, or RAG, one is injecting content that has been retrieved from existing documents. CAG is like an infinite extension of RAG, in which an infinite amount of content can be generated on the fly\u2014using computation\u2014to feed to an LLM. Internally, CAG is a somewhat complex piece of technology that has taken a long time for us to develop. But in its deployment it\u2019s something that we\u2019ve made easy to integrate into existing LLM-related systems and workflows. And today we\u2019re launching it, so that going forward any LLM system\u2014and LLM foundation model\u2014can count on being able to access our Foundation Tool, and being able to supplement their capabilities with the superpower of precise, deep computation and knowledge. <\/p>\n Today we\u2019re launching three primary methods<\/a> for accessing our Foundation Tool, all based on computation-augmented generation (CAG), and all leveraging our rather huge software engineering technology stack.<\/p>\n For further information on access and integration options, contact our Partnerships group \u00bb<\/a><\/em><\/p>\n <\/a><\/strong> <\/a><\/p>\n","protected":false},"excerpt":{"rendered":" … to LLMs access to Wolfram tech<\/span> in a standard, general … the Wolfram Language. The Tech<\/span> to Use Our Foundation Tool … the foundation tool that our technology<\/span> represents. And the good … our rather huge software engineering technology<\/span> stack. Immediately call our …<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"","fifu_image_alt":"","footnotes":""},"categories":[1],"tags":[],"class_list":["post-100061","post","type-post","status-publish","format-standard","hentry","category-news","wpcat-1-id"],"_links":{"self":[{"href":"https:\/\/new7.shop\/zerocostfreehost\/index.php\/wp-json\/wp\/v2\/posts\/100061","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/new7.shop\/zerocostfreehost\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/new7.shop\/zerocostfreehost\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/new7.shop\/zerocostfreehost\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/new7.shop\/zerocostfreehost\/index.php\/wp-json\/wp\/v2\/comments?post=100061"}],"version-history":[{"count":0,"href":"https:\/\/new7.shop\/zerocostfreehost\/index.php\/wp-json\/wp\/v2\/posts\/100061\/revisions"}],"wp:attachment":[{"href":"https:\/\/new7.shop\/zerocostfreehost\/index.php\/wp-json\/wp\/v2\/media?parent=100061"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/new7.shop\/zerocostfreehost\/index.php\/wp-json\/wp\/v2\/categories?post=100061"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/new7.shop\/zerocostfreehost\/index.php\/wp-json\/wp\/v2\/tags?post=100061"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}The Practicalities<\/h2>\n
Immediately call our Foundation Tool from within any MCP-compatible LLM-based system. Most consumer LLM-based systems now support MCP, making this extremely easy to set up. Our main MCP Service is a web API, but there\u2019s also a version that can use a local Wolfram Engine<\/a>.<\/p>\n
A one-stop-shop \u201cuniversal agent\u201d combining an LLM foundation model with our Foundation Tool. Set up as a drop-in replacement for traditional LLM APIs. <\/p>\n
Direct fine-grained access to Wolfram tech for LLM systems, supporting optimized, custom integration into LLM systems of any scale. (All Wolfram tech is available in both hosted and on-premise form.)<\/p>\n
Wolfram Foundation Tool Capabilities Listing<\/i> \u00bb<\/a><\/em><\/p>\n