Where Are We? The AI Universe

Words such as AI and machine learning has been circulating vivaciously among the investor and founders community alike, with buzzwords such as large learning models, entering our parlance at an intimidating velocity. But what is deep learning exactly, how is it related to computer vision, and where do large language models fit into the picture?

Below, we take a quick look at the AI universe by exploring the different model types, what applications each type of learning model can be utilized on, and where future trends may lie.??

1. Buzzwords, AI slang, and key terms:??

Artificial Intelligence (AI)- artificial intelligence is leveraging synthesis of large datasets and compute power to make inferences from given inputs; it is a large superset of all different types of modeling, defined below:

• Machine Learning (ML) - machine learning includes making inferences via pattern matching from data without explicit programming; types of ML models include supervised learning, unsupervised algorithms, and deep learning.
• Natural Language Processing (NLP) - utilizes datasets of texts to interpret or generate new text. Model types include content extraction, language translation, and the exciting area of large language models (see below).
• Computer Vision - algorithms that derive information from images or videos; common examples include convolution neural networks or algorithms used to identify moving objects such as cars, humans.?
• Large language models (LLM): subset of NLP- refers to large algorithms leveraged on cosmic datasets that can read, summarize, and generate texts akin to how humans write; most recently entered public conversation via the explosive release of ChatGPT and BERT.

2. Major applications in mainstream tech (by no means exhaustive) list:

• Computer vision is used in self driving cars (Waymo, Rivian), detection of abnormal medical imaging (Path AI);??
• Machine Learning is broad in application and leveraged in classification, deep learning (Deepmind, Netflix, Spotify)
• Natural Language Processing is used in voice recognition, voice assistants (Alexa, Siri, Infinitus*), processing text, generative text (ChatGPT, BERT)

3. Exciting frontiers

We at Tau continue to double down on our central thesis revolving around AI, particularly in its applications in healthcare, automation, and robotics. As frontiers in algorithm are continually pushed forward, we see the following key macro trends starting to emerge and become established:

• The ethos of AI has revolved around accessibility and open source, where large research organizations such as OpenAI were started as a nonprofit. We believe that ethos will continue as major innovations in other areas of artificial intelligence blossom. Just as platforms Tensorflow and Kaggle enabled the first wave of accessibility for commoditization of machine learning algorithms, we believe that future revolutions of AI will be best served through democratization of access, incentivized on both an ethical and competitive advantage front.
• Previous limitations such as computational power, speed, and scalability have been incrementally tackled. However, industries where data is siloed or inaccessible, such as healthcare, will be largely unaffected by innovations in this front- the emphasis of data interoperability and availability will be even more apparent as a barrier for technological adaptation when compared side-by-side with other industries that will explode in speed and compute. We challenge founders to solve problems in and surrounding data security, accessibility, and sharing just as much as they focus on algorithm choice when leveraging AI.?

Primary author of this article is Sharon Huang. Originally published on “Data Driven Investor.” ? *disclaimer: Infinitus is a portfolio company of Tau. These are purposely short articles focused on practical insights (we call it gl;dr — good length; did read). See here for other such articles. If this article had useful insights for you, comment away and/or give a like on the article and on the Tau Ventures’ LinkedIn page, with due thanks for supporting our work. All opinions expressed here are from the author(s).