A.I. Executive Briefing #1

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

1. Global regulatory response to A.I.
2. Marc Andreessen jumps into the Pro A.I. vs Anti A.I. Debate
3. Leaked OpenAI Plans
4. 30min MRI scans detect Cancer with A.I.
5. Nvidia 2D to 3D Image Model
6. A.I. causes almost 4k jobs to be lost in May

VENTURE FUNDING

1. Content moderation startup Unitary raises $8M Series Seed
2. LLM data framework startup LlamaIndex raises $8.5 Series Seed
3. LLM search startup Glean raises $100m in a Series C
4. A.I. startup Cohere raises $270M in a Series C

News Round-up

1. EU & US Regulation, Australia vs Japan & India

Early in May, we saw Sundar and Altman go on tours in Europe to court law makers and attempt to broker deals with them regarding future legislation. Additionally, Anthropic CEO Dario Amodei and Microsoft president Brad Smith were present at the U.S.-EU Trade and Technology Council (TTC) which took place in Sweden at then end of last month. As of know, we know a few things:

1. No legislation has been passed, but the drafts are in the works. Private sector and public sector stakeholders alike are clamoring to get their word in edgewise on how strict or relaxed regulation should be, but one thing is for certain, everyone is insistent on regulation (well almost everyone)
2. US Secretary of state for commerce stated that “it will take a little bit of time for the US Congress or the parliament or other regulatory agencies to catch up” to the current state of A.I. development.
3. Audits of A.I. models, alignment processes, and outputs are likely to be the stop gap while the regulators play catch up.
4. Different countries have different opinions on how copyright law should be enforced regarding A.I. training data. Australia taking a stance of enforcing copyright laws on A.I. model training data while Japan has explicitly stated it will not enforce copyright laws on A.I. model training data.

2. Marc Andreesen vs AI Doomers & Conventional Media

Marc Andreesen takes a strong point of view against the incessant A.I. doomerism that has been proliferating throughout conventional media and seeping into the perspective of the general public. With several open letters and claims to the potential end of humanity, Marc has laid out five counter points to what he deems an anti-A.I. Cult and one concession to the opposition.

1. A.I. will not kill us all.
2. A.I. will not ruin society and is more likely to increase our societal productivity.
3. A.I. will not take all of our jobs. This technology will result in increased output with the same or fewer inputs, which will drive demand to other parts of the economy increasing the need for jobs overall. At the most there will be a reshuffling of jobs resulting in a net need for more work.
4. A.I. will not lead to crippling inequality as the trend of technology is democratization. This notion is supported by the fact that the A.I. movement is primarily open source.
5. A.I. will enable bad people to do more bad things just as it will enable good people to do more good things. He suggests an investment in “Defensive” A.I. technology. He states, “Put AI to work in cyberdefense, in biological defense, in hunting terrorists, and in everything else that we do to keep ourselves, our communities, and our nation safe.”

3. OpenAI’s Plans that were leaked and taken down:

OpenAI’s internal plans were “leaked” in an article and taken down (but we got it). In the article, it mentions that this year their priority is to make GPT-4 faster & with cheaper compute. Longer context windows are a must, but unfortunately they are limited by a lack of GPUs. Their parent company is hoping that their partnership with AMD will eventually help stave off their dependency on NVIDIAs GPUs that seems to be OpenAI’s current bottle neck. In 2024, multimodality will be a key focus as this was originally slated for GPT-4s release but was held back due to do GPU constraints.

4. Ezra A.I. CEO launched Ezra Flash AI, enabling world’s first 30-minute full body MRI.

Ezra A.I. was founded in 2018 and was focused on implementing A.I. in order to reduce the time MRI scans take. Five years later they have reduced the time of an MRI from 90+ minutes down to 30 minutes available for around $1300. Utilizing an exclusive collection of more than 5,000 MRI scans, Ezra utilizes this longitudinal dataset to detect bodily transformations at an early stage. Additionally, the system autonomously generates reports incorporating the scan results and interpretations from radiologists, which are sent to healthcare experts for fine-tuning and review. This automation enables Ezra to deliver reports to its members within 3-5 days instead of the previous 7-10 days. With the ability for radiologists to see more patients and potentially screen individuals earlier for diseases like cancer, an immense amount of personal value can be delivered to individuals and an immense amount of notional value can be saved in our healthcare system.

5. Neuralangelo: NVIDIA x JohnHopkins releases High fidelity 3D Surface model

Current methods struggle to recover detailed structures of real-world scenes. Even without auxiliary inputs such as depth, Neuralangelo can effectively recover dense 3D surface structures from multi-view 2D images with fidelity significantly surpassing previous methods, enabling detailed large-scale scene reconstruction from RGB video captures. With a lot of innovation in the Generative AI field (midjourney & stable diffusion) and specifically related to text-to-image (Style drop by google being the most recent), there is potential for a Text to 2D image to 3D model. Models that can be utilized as 3D assets in CAD models for 3D printing can turn text on a computer directly into a physical object in real life.

6. Challenger Report on Job Cuts due to A.I.:

A Challenger Report was released showing that U.S.-based employers announced 80,089 cuts in May, a 20% increase from the 66,995 cuts announced one month prior. The Technology sector announced the most cuts in May with 22,887, for a total of 136,831 this year. Of the 80k jobs cut ~ 4k were cited to be as a direct result of Artificial Intelligence (4.8%). This trend will be an interesting one to watch especially as we continue to see the limitations of A.I. applications while also pushing the limits of the tech.

Funding News

7. Content moderation startup Unitary raises $8M Series Seed

Unitary is launched an open-source tool, Detoxify, which has already been downloaded over 7 million times. The funding round will be utilized to build out their team and continue to refine their Detoxify repository.(Link)

8. LLM data framework startup LlamaIndex raises $8.5 Series Seed

LlamaIndex allows you to seamlessly integrate individual or enterprise data, including files, workplace apps, and databases, with LLM applications. Born out of GPTIndex launched by founder Jerry Liu in November of 2022, the project reached over 20k stars on github and over 200k downloads 6 months after launch. (Link)

9. LLM search startup Glean raises $100M Series C at $1Bn valuation

Glean claims to deliver a “unified search experience” to all of a company’s employees by connecting all of your internal tools to Glean. What’s crazy is that they came out of stealth less than a year ago and just raised $100M. Smells fishy.. (Link)

10. A.I. startup Cohere raises $270M in a Series C

Toronto-based Cohere is valued at over $2Bn with this new round of funding from a mix of venture capital and strategic investors including Oracle Corp. and Nvidia Corp (Link)

Reader Questions

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.