No. 210: Conversational AI | Number Sense | Walking

Speech and language: the crown jewel of AI with Dr. Xuedong Huang (Microsoft Research)

This interview and the next one are examples of this shift picking up speed in the development of artificial intelligence where we’re not just feeding data to computers as part of training it for one task. The computers are drawing insights from multiple angles and engaging new information in increasingly complex ways.

A new AI acquired humanlike ‘number sense’ on its own (Science News)

Why Walking is the Key to Being More Productive (GQ)

As the weather gets warmer, my urge to wander increases. You won’t catch my in an arctic pole, but if you’re driving around DC late at night, there’s more than a slight chance you’ll see me strolling.

No. 209: The Rihanna Force | Brain Imaging AI | Space AI

Why Rihanna Broke Barriers That Others Couldn’t (Trapital)

I’m looking forward to seeing what becomes of this joining of forces with LVMH. Dan Runcie, as usual, does a real nice job breaking things down.

Brain imaging AI leader icometrix raises $18 million in new funding (icometrix)

This is a very interesting company. My grandmother has Alzheimer’s and I have two aunts with various levels of dementia. I look forward to a day when technology gives Alzheimer’s and dementia a run for their money.

Hypergiant uses satellites and AI to mine oil and gas data (VentureBeat)

Hypergiant presents an interesting use case for machine learning, layering it on top of satellite data to uncover insights for its clients. The oil and gas exploration support has me wondering what the applications could be for improving the security of pipelines in Nigeria and elsewhere.

No. 208: AI Drives News and Music | Alibaba’s Global Ambitions | 2019 Fundraising Environment

TikTok Is the New Music Kingmaker, and Labels Want to Get Paid (Bloomberg)

I wonder what it is about the music industry that has it looking at itself on the wrong side of a deal once again. Decades ago, Napster and file sharing put a massive dent in the industry, then Apple came along and dictated the price it was going to pay for music. Now, it finds itself trying to figure out how to get value out of the virality that TikTok is driving among artists.

What’s more interesting to me is the engine behind TikTok. The app is one of the products of Bytedance, the most valuable startup in the world at $75B. Bytedance owns a number of mobile-first products that are all powered by machine learning. There’s a news app Toutiao that is powered by artificial intelligence technology that can write 400 words in a couple seconds. There are 140 million monthly users spending around an hour a day on that app in China. Incredible.

I wonder what the quality is of the news the app is developing. To what extent is Bytedance in coordination with the Chinese government? What sort of unintended insights is the company getting as a result of having its finger on the pulse of people’s music and news consumption? How much more room for growth is there in its valuation before it goes public?

Breaking down Alibaba’s global ambitions (Digiday)

I’m yet to buy anything through Alibaba, but it seems like those days are more and more numbered as the company expands its footprint globally. The amount of handholding that its various subsidiaries engage in to ensure the success of brands is pretty remarkable.

The Fundraising Environment in 2019 – Three Major Shifts (Tomasz Tunguz)

Tom Tonguz argues that the fundraising environment for startups is very sophisticated in three ways – the types of funding founders can pursue, the sophistication of the pricing, and the speed with which money can be raised.

I agree with his takes on the first and third point, but wonder whether the pricing is really all that sophisticated. Investors are using enterprise value to forward revenue to value companies. Are startups hitting their forward revenue targets? If they aren’t, what adjustments are investors making to their valuations? How are late stage investors remaining disciplined in the face of Softbank dropping $1B in a startup’s bank account?

No. 207: China’s AI Exports| A Historian Bot | Trustworthy AI Framework

AI, Foreign Policy, and National Governance Impact: Focus on China (Modern Diplomacy)

Arthur Gwagwa outlines the potential bad effects of China’s inroads throughout the Global South and export of its AI technology particularly to African countries. I share his concern that these technologies could help bad leaders on the continent dig their heels in through ratcheting up a surveillance state.

I disagree with his view that the West should come in and save African nations and other members of the Global South from China’s incursions. I posted yesterday about how the U.S. doesn’t have an AI strategy. Despite US attempts to isolate Huawei, Britain is going ahead to contract with the company to build 5G networks. How are these countries going to counter China in other countries without a strategy and being on the same page?

African countries and other members of the global south have to figure out how to avoid yet another principal-agent table tennis match and figure out how to be stronger and stronger agents on a global scale in shaping AI and its uses.

Metahistory for (Ro)bots: Historical Knowledge in the Artificial Intelligence Era (Historia da Historiographia)

This is a very interesting paper that anticipates the development AI-powered historian bots that serve as our future textbooks on world history. The authors propose a method for developing that technology well, in efforts to avoid any bad effects.

Until the lion tells his side of the story, the tale of the hunt always glorifies the hunter.

This quote was ringing in my head while working through the paper. The authors outline how the data sets containing all the historical information need to be structured properly so they can be scanned into the database the bot would draw from well. What happens to cultures that still rely on oral history? What about parts of the world where everything hasn’t been digitized yet?

I could see a historian bot being an incredibly powerful tool to surface history that is typically glossed over, but the inputs have to be broad ranging and accurate. I could also see a historian bot baking in poorly informed historical takes that shape biases and lead to the kind of ignorance we see in the world today

Establishing the rules for building trustworthy AI (Nature)

It is time to ‘make haste slowly’ (festina lente) in the development of AI.

Luciano Floridi, a member of the High Level Expert Group put together by the European Commission, outlined a framework for how the EU should shape the development of AI by putting the brakes on its development and asking ethics questions early on.

The amount of compute in AI processes is doubling every three and a half months. I don’t see how the folks building these technologies work well with stakeholders like the High Level Expert Group when the group took a year to develop a 7-point framework for building trustworthy AI. The technology is advancing too quickly.

No. 206: Congressional AI Task Force | NYC AI Center | Microsoft Opens AI Playbook

Waters Announces Committee Task Forces on Financial Technology and Artificial Intelligence (US House Committee on Financial Services)

I’m glad to see Rep. Alma Adams on the artificial intelligence committee. My hope is that she asks those questions that get at the unintended consequences around the meshing of AI technology and financial services and those impacts on vulnerable Americans.

The concerning part about this is that this is work that should have taken place a years ago. The Chinese government released its AI strategy in 2017 and cities are investing in AI research at the level of the U.S. government. The White House just started thinking through AI policy a couple years ago. It’s water under the bridge now, but I look forward to following this committee and hope they staff up well and move quickly.

Silicon Valley likes to think of itself as successful on its own, but those defense contracts 40 years ago served as a huge wind to its back. Even today, companies like Tesla have benefited greatly from government investment. Currently, the private sector is investing heavy in AI. The US won’t lead in this space until the government gets itself together to put resources behind the development of AI technology in a coordinated fashion.

As an aside, I remember seeing Rep. Adams around Greensboro as a kid growing up. She and Mayor Pro Tem Yvonne Johnson shaped my view of who could lead. Grateful for their model.

(Erica, I’m rooting for you!)

NYCEDC Seeks Proposals for Responsible Artificial Intelligence Center (GlobeSt)

Cities across the US would do well to follow New York City and have at least a few folks think proactively about AI and how their city can shape its development and impact on their locale. This effort by NYC to stand up a center that shapes the development of artificial intelligence is important and commendable.

There’s one caveat. I mentioned in the above section that cities are investing in AI at the level of the U.S. For more context there, Shanghai is standing up a nearly $15B fund to invest in AI research and development. Contrast that with the $7M with which NYC will support its Center for Responsible Artificial Intelligence.

Creating AI glass boxes – Open sourcing a library to enable intelligibility in machine learning (Microsoft Research)

This is a strong move by Microsoft to build a tool that helps folks understand what the AI they are using is actually doing. The movie Margin Call came to mind while reading this. Something like this might have been helpful as banks were spinning up models they didn’t understand in the lead up to the financial crisis of 2008.

No. 205: Jumia a Fraud? | Black VCs Get That Princeton Bag | Uber Creator Secures $3.7B

Not All IPOs are Created Equal. Jumia is a Fraud. (Citron Research)

This tea is hot, but I’m not sure there’s enough honey in this report. Jumia has earned most of the side-eyes it gets but I don’t think this report hits the bar for a fraud claim.

Princeton Looks to Break Up the White Male Money Monopoly (Bloomberg)

This title gives Princeton too much credit. Nonetheless, I’m excited for 645 Ventures securing them as an institutional investor.

How Much Is an Idea Worth? In Uber’s Case, $3.7 Billion (Bloomberg)

This is a prime example of why the trope, “Keep your ideas to yourself. You don’t want folks to steal them,” is BS.

No. 204: Morning Reads

Creative Stock: Why Intellectual Property is Africa’s Most Valuable Resource (Medium)

Losing $2m in Nigeria’s music industry (Medium)

The failure of Reconstruction was a ruthless act of sabotage (Washington Post)

Nipsey Hussle had a plan to beat gentrification — in South L.A. and across the U.S. (LA Times)

Why Ramaphosa is probably not in a position to end corruption and patronage (Daily Maverick)