Stack Overflow, Shopify, LinkedIn, Uber inject A.I. into their products

The A.I. Executive Briefing is an expert weekly curation of A.I. news by our research team, shared externally now because we feel there’s too much hype & noise in the market. The same content will be distributed through?this substack .

NEWS ROUND-UP

I. Product

1. Stack Overflow Strikes Back

2. New Shopif-AI Product: SideKick

3. Rap your head around Google’s TextFX

4. Linkedin Links up with AI

5. Uber joins the A.I. arena

II. Regulation

6. US vs China: To Export AI or to Import AI

III. Research

7. IBM Watson on the Map?

VENTURE NEWS

8. AI chip startup Tenstorrent lands $100M investment from Hyundai and Samsung

9. GoStudent adds another $95M to its war chest to go after VR and AI-enhanced tutoring

10. RapidAI Raises $75M in Series C Funding

11. Inworld AI raises $50M round at $500M valuation for AI game characters

12. Akkio Secures $15M in Series A Funding to Accelerate AI Platform Commercialization

13. Thymia Raises $2.7M in Seed Funding

14. Akhetonics raises €2.3 million for its all-optical processor prototype

NEWS ROUND-UP

I. Product Updates

1. Stack Overflow Strikes Back

In what seems like a response to all the negative press Stack Overflow received over a drop in their website traffic dropping , the company has announced Overflow AI . A couple of highlights here: Integrations (with Slack, Teams & VScode) so users can ask questions directly to Overflow AI, users can curate and build a unique knowledge base from proprietary data, & most importantly they launched the new Discussions feature.

Stack Overflow is built on conversations between engineers seeking help & engineers with the skills to provide answers. Discussions help stimulate and grow the forum of discourse that Stack Overflow has built its website on.

Why it matters: We’ve stated our viewpoint many times in the past that proprietary data will be the most valuable asset in this LLM landscape for the foreseeable future. While it’s great that Stack Overflow seems to be training a model on their “proprietary data” we know that all of the major model providers like OpenAI have already scraped all of this data as it’s publicly available on the open web. Their drop in traffic has been attributed to LLM applications like ChatGPT and Github Copilot solving these same questions that users used to submit to Stack Overflow communities. However, we know that communities are highly valuable and extremely difficult to replicate so we’d be interested to see how SO leverages this going forward as they struggle to compete with competition.

2. New Shopif-AI Product: SideKick

Sidekick is a conversational AI assistant that’s trained to “know and understand all of Shopify,” and can offer as an assistant for those building the websites. Shopify Magic can provide answers tailored to conversation histories & policies, and generate blog posts, product descriptions and marketing email content. With over 2.1m DAU , it is fair to say that Shopify is determined to make itself the easiest platform for creators to build maintain and market their e-commerce products.

Why it matters: This just makes sense and is a good move for Shopify to leverage AI to better support their existing customers and make the application fine-tuned to Shopify’s data + use case. Shopify has done a good job lately of staying relevant in the A.I. arena (e.g. partnership with OpenAI) and also with displaying corporate discipline during these times (e.g. releasing a “meeting calculator” that calculates the cost of a meeting based on the salary of the attendees). The market seems to agree with these decisions as SHOP is up 58% YTD.

3. Rap your head around Google’s TextFX

Google, in what seems like an avalanche of AI product releases , has recently released a new LLM focused on creatives using language in partnership with renowned Hip-Hop artist Lupe Fiasco. TextFX itself breaks down the elements of language into about 10 different aspects to allow a unique level of control over language.

This system gives a powerful exploratory tool to assist creators during their creative process. This is a suite of tools built to be a specific application for creators. From exploring new forms of language using AI to generating new and original sound using AI, the writer or soundsmith is put at the center of the process only leveraged to a new level with these new tools. Here’s a short youtube video where Lupe Fiasco explains TextFX.

Why it matters: The TextFX playground is quite interesting if you go to their website. I wonder how many YC companies are now destroyed because of this release… or perhaps how many YC companies will now be created with this release? I think there’s an interesting philosophical debate to be had around where the boundary is (if there even is one) of what creatives would deem as true art if A.I. helped generate some or all of the content.

4. Linkedin Links up with AI

One of the common use cases of LLMs is to hone it into a coach. Linkedin is building on its professional brand and is planning on launching a career coach! This AI use is very different from the profile optimization they announced earlier this year. The LinkedIn AI coach seems to be an assistant for job application, company selection, and course selection & offers new ways to connect with those in your network.

Although these applications are not new in themselves, web-based applications are hesitant now to let LLMs from other companies crawl all through their data and have opted to launch their own products instead, specifically built on data relevant to the end user (we can only hope).

Why it matters: Similar to Shopify releasing their AI Sidekick, using AI to enrich the current user experience for LinkedIn users is a no-brainer. However, LinkedIn has extremely rich proprietary data on user profiles, and so they have the potential of creating something truly ground-breaking IF they were to leverage AI on their dataset.

5. Uber joins the A.I. arena

Uber, which recently just reached profitability after over a decade of operations, is joining its competitors in a league of mobile-based transportation/delivery LLMs. Uber , DoorDash , and Instacart are all in the process of releasing their own AI tools and chatbots for their users. It is obvious that of all the chatbots that will be released only a handful will capture meaningful adoption from retail users.

The behavior of people to change from using an app like Uber or DoorDash to order food or a car to talk to a bot will be hard to incentivize and even tougher to argue. The core interaction with LLMs brought up by the initial rollout of the chatGPT plugins page was having an LLM draft a grocery list and then execute the order in Instacart or plan a vacation and execute a car scheduling in Uber, not for an end user to start chatting with a bot in the app itself. It has yet to be seen how these apps will drive us toward maximal efficiency or insanity.

Why it matters: I find the creation + integration of LLM-based chatbots for these ridehailing companies to be a bit different than my support for the AI released by Shopify + LinkedIn. As opposed to the user experience of Shopify + LinkedIn, these ride-hailing apps are designed and optimized for the least amount of clicks you have to make on your phone. Besides a customer support bot use case, I fail to see how integrating some type of LLM-bot will improve the UX.

II. Regulatory Updates

6. US vs CHINA: To Export AI or to Import AI

In a battle of regulatory Ping Pong, China, and the US have been passing executive orders and laws that will make it increasingly harder for businesses up and down the AI stack and across the Pacific to do business with each other. The executive order currently in question is slated to be signed any day now and will go into effect next year to limit critical US technology investments in China. Recently, China required licenses to export 2 key metals at the beginning of this month in response to the US banning cloud computing usage and certain key CPUs and GPUs exports to China. It is even being reported that NVIDIA chips are going for as high as $70k in China for Nvidia's H800 compute GPU.

All of this activity between the US, China, and NVIDIA has inspired other chip manufacturers to capitalize on the supply vacuum artificially created by the government squabble. AMD has planned a China GTM for its newest chip to compete with NVIDIA to capture the growing global demand for GPUs. It is worth also noting that Intel and NVIDIA also both have designed chips to be sold directly to China that fall within the US’s newest regulatory architecture regarding this topic.

Why it matters: This is all about cascading effects, or 2nd/3rd order consequences. At the moment, the U.S. has more leverage than China as they take regulatory jabs at each other, capitalizing on the global AI momentum. However, we see that these gaps created by these regulations are quickly filled by the market due to the pure nature of capitalism. What cascading effects these geopolitical tensions have on US and China respectively is yet to be seen.

III. Research Updates

7. IBM Watson on the Map?

NASA has teamed up with IBM using their latest Watsonx model to create the largest opensource geospatial transformer model currently hosted on Hugging Face. This model is the first open-source AI foundation model released by NASA and it can analyze geospatial data up to four times faster and with half as much training data than state-of-the-art deep-learning models, IBM has?estimated .

The model is Multi-Modal and can ingest text, digital data, satellite images, and other weather indicators to inform its predictions.

Why it matters: IBM Watson is finally on the map again…where have they been?

VENTURE NEWS

8. AI chip startup Tenstorrent lands $100M investment from Hyundai and Samsung

Amidst all of the mania regarding NVIDIAs valuation, China and the US engaging in this high-tech cold war & the general demand swell for high-end computing chips, and more new entrants are attempting to capture markets to provide GPUs or cloud computing infrastructure to those in need of this technology.

Founded by an ex-AMD team, Tenstorrent sells AI processors and licenses AI software solutions and IP around RISC-V, the open-source instruction set architecture used to develop custom processors for a range of applications. Additionally, the company has an AI DevOps product, allowing teams to train AI models using cloud computing prior to acquiring hardware. This round was a convertible note round , which obfuscates the valuation. Tenstorrent last raised $200 million at a valuation eclipsing $2 billion.

9. GoStudent adds another $95M to its war chest to go after VR and AI-enhanced tutoring

GoStudent?is a tutor marketplace and has raised a strategic $95 million in a mix of debt and equity from Deutsche Bank and other investors, including Left Lane Capital, DN Capital, Tencent, Prosus, DST, Coatue, and Softbank Vision Fund 2. With lots of corporate VC and another “counteroffensive” investment from Mayoshi Son , GoStudent plans on using the funds to entertain an AI lesson plan generator based on the curriculum from their 23k tutors who use their marketplace.

With a community of educators creating original content to assist students with learning a variety of topics, an LLM trained on the specific needs of their users, specifically to assist tutors in teaching is a strong use case. It is important that companies at least attempt to build out AI product features to prevent users from unintentionally ignoring their platform. Not dissimilar to Stack Overflow, preserving online communities and the natural original content generated from users using the website is probably the most important thing for these companies to protect.

10. RapidAI Raises $75M in Series C Funding

The developer of Artificial Intelligence (AI) and technology workflow solutions to combat life-threatening neurovascular, cardiac, and vascular diseases, raised $75m in Series C funding led by?Vista Credit Partners, a subsidiary of?Vista Equity Partners. Their solutions focus on a variety of disease states and generating more FDA-approved (more as in they already have several) AI systems or AI-enabled devices to assist in diagnoses will only further cement themselves as a competitor with Incumbents like Google attempting to push their Med-PaLM 2 model as best as they can.

11. Inworld AI raises $50M round at $500M valuation for AI game characters

Inworld AI is focused on generating high-fidelity intelligent NPCs within video games. This involves not only personality modeling but also character modeling. The Inworld Character Engine powers multimodal character expression by orchestrating multiple machine-learning models designed to mimic the human brain and communication. These NPC “brains” can be linked to 3D character models and deployed in-game. They generate unique knowledge and remember their experiences. The company raised $50 million in funding from?Lightspeed Venture Partners , bringing the valuation of the maker of an AI-based character engine for games to over $500 million.

Although this seems like a highly disguised excuse to develop closed-loop AI agents, the developments here are critical. Already featured in Unity’s new AI app marketplace, Inworld is actually pioneering the application of AI agents within closed environments. The question still remains: Will AI agents be usable by laymen or if the tech will remain only for the most technically experienced?

12. Akkio Secures $15M in Series A Funding to Accelerate AI Platform Commercialization

Akkio offers a no-code platform for enterprises to deploy artificial intelligence. The platform aims to simplify AI development and deployment by using no-code tools that users can build and deploy AI systems for tasks such as churn reduction, fraud detection, and sales funnel optimization. They raised $15 million in a series A funding round led by Bain Capital Ventures and Pandome, Inc.

Companies like Salesforce and Amazon skipping the “Build a unique own LLM” step in the near term (Amazon is still working on a variety of LLMs) and focusing instead on the “Use other LLMs to build on our platform,” there is definitely going to be some competition here. Currently, the go to place to building with AI is Replit, so Akkio taking a no-code approach may work while people are still figuring out the best combo of models to use case and new models/updates seemed to get pushed out almost daily. In the long term Akkio will need to either lock in their acquired users or find a fresh niche of enterprise clients to target.

13. Thymia Raises $2.7M in Seed Funding

AI systems have been getting a lot of attention for cancer detection, early diagnosis & radiologist/MRI assistance. Thymia is looking to pioneer a mental illness diagnoses AI system to replace current survey-based methodologies of diagnosing mental illnesses. The company has raised $2.7m from Kodori Ventures.

They have designed a proprietary model that listens to people’s voice & watches their body language to screen for anxiety & depression. They trained their model on over 6k patients, arguably one of the largest datasets related to depression. They released a game that patients can spend 10 minutes playing and the results are sent to a clinician. Currently a pureplay SaaS tool with all its features for clinicians, the team found that psychologists and doctors were more keen to use modules, in the form of specific APIs and gamified widgets, as opposed to the entire platform. Being aware of how their users are utilizing their platform and where they prefer to spend most of their time is critical to maintaining product market fit in a competitive and emerging market such as healthcare and AI diagnosis.

14. Akhetonics raises €2.3 million for its all-optical processor prototype

Akhetonics claims to have multiple prototypes of the optical transistor and the foundation of a processor’s design. All-optical processors will have higher bandwidth and faster speeds with greater efficiency and information density than electronic ones. While lots of attention is still being placed on silicon chips, especially with the rumors of a potentially frictionless silicon chip floating about, Akhetonics is focused on building a chip that uses is inherently frictionless— light. The company is aiming for a full-scale optical CPU, with a full prototype by 2024. Although new designs are critical to diversifying away from the silicon chips that currently dominate the high-performance computing landscape, manufacturing capabilities and upstream solutions are still critical to alleviating the supply-demand disparaty that exists today.

Send us a message with any questions/comments/thoughts on anything A.I. related and we’ll try to answer them in our next release.