⚙️ A different kind of legal AI

November 6, 2024

Good morning. As of late Tuesday night, the U.S. presidential race was tight, far too tight to call. The votes keep on rolling in, but the counting will likely continue through the next few days.

As always, we shall see what happens.

— Ian Krietzberg, Editor-in-Chief, The Deep View

In today’s newsletter:

🛰️ AI for Good: Space lasers and space junk
💻 Anthropic raises the price of its new model
🏛️ Big Tech and the US election
👨‍⚖️ A different kind of legal AI

Source: West Virginia University

Hard though it may be to believe, the ‘space’ around our planet is quite crowded.

In orbit around the Earth are tens of thousands of active and inactive satellites and more than 36,000 pieces of general space debris, according to the European Space Agency. And that’s only the junk that the ESA tracks. In total, there are close to 13,000 tons of mass orbiting Earth.

The problem: The issue with this is simple: we have a lot of important infrastructure up in orbit. And if that important infrastructure can’t dodge these thousands of chunks of debris in time, things could get damaged, impacting us on the ground (think WIFI networks, GPS, etc.).

The problem has been recognized by many authorities, including a former NASA administrator I spoke to in March.
NASA is trying to figure out ways to address the debris; some have proposed distributing literal nets to do the work.

A solution: Hang Woon Lee, a mechanical and aerospace engineering professor at West Virginia University, is working on an AI-powered system to protect our orbiting infrastructure.

It involves lasers.

Though the work is in its infancy, Lee’s vision is the creation of a cost-efficient network of lasers, mounted to active satellites and powered by artificial intelligence.
This approach — which has been funded in part by NASA through a research grant — is specifically designed to protect satellites from smaller pieces of debris, rather than large ones.

Instead of destroying the debris in a Star Wars-esque display of technology, the intention of the lasers here would be to simply nudge the pieces of debris out of the way.

Best Buy has a knack for picking the up-and-coming tech products that go on to dominate the market. Their early bets on household items like Ring (acquired by Amazon for $1.2B) and Nest (acquired by Google for $3.2B) have a proven record of paying off.

Now Best Buy is lifting the curtain on their latest find, launching RYSE’s SmartShades in over 120 retail stores. RYSE has already hit $9M+ in lifetime revenue with over 60,000 units sold, and the numbers are rising (along with the window shades).

RYSE shareholders have seen their value increase 40% year-over-year, with strong upside remaining as they scale into retail and high-volume B2B channels.

Invest in RYSE at just $1.75/share before it becomes a household name.

Source: Anthropic

When Anthropic unveiled Claude 3.5 Haiku last month, the company suggested that it would cost the same as its predecessor, Claude 3 Haiku. But in a post this week, the company said that the model would instead cost more.

The details: The upgraded 3.5 Haiku model “matches the performance of Claude 3 Opus, our prior largest model, on many evaluations at a similar speed to the previous generation of Haiku,” according to Anthropic.

It will cost $1 per million input tokens and $5 per million output tokens.
In comparison, Claude 3 Haiku costs $.25 per million input tokens and $1.25 per million output tokens, a 4x increase.

The reason behind the price increase, according to Anthropic, has to do with the model outperforming Claude 3 Opus “during final testing.”

Why it matters: The promise of generative AI is one of efficiencies. Right now, developers are subsidizing the enormous cost of training and operating their generative AI models in order to get people on board.

But I have said many times that this approach cannot continue forever; at some point, the developers — and the venture capitalists backing them — will want to see a return on their investment, and so the costs will eventually be passed onto the users, either through higher subscription fees or through advertising.

This is likely the first price increase of many. And it will begin to stress test the willingness of the user base to pay for access to these tools; at a certain point, the cost of generative AI for the users could prove to be higher than just hiring humans at minimum wage.

This free 3-hour Training on AI & ChatGPT (worth $399) will help you become a master of 20+ AI tools & prompting techniques and save 16 hours/week.

Get it now for absolutely free! (for first 100 users only) 🎁

You will learn how to:

Build business that make $10,000 by just using AI tools
Make quick & smarter decisions using AI-led data insights
Write emails, content & more in seconds using AI
Solve complex problems, research 10x faster & save 16 hours every week

Register & save your seat now! (100 free seats only)

Pipedrive is one of the best CRMs out there. Plus, their AI-powered sales assistant sharpens sales teams, significantly boosting performance.
- Level up your sales success with Pipedrive today.*
Robotics startup Physical Intelligence raised $400 million in funding, led by Jeff Bezos and with a contribution from OpenAI.

Live updates for the US Election (The New York Times).
Amazon gets FAA approval for new delivery drone as it begins tests in Arizona (CNBC).
How the world is viewing the US presidential election (Semafor).
Inside the plan to use AI to purge voter rolls (404 Media).
Perplexity nears $9 billion valuation in new investment round (The Information).

If you want to get in front of an audience of 200,000+ developers, business leaders and tech enthusiasts, get in touch with us here.

South Korea’s privacy authority on Tuesday fined Meta $15 million for illegally collecting sensitive personal information — including information regarding political views and sexual orientation — from Facebook users, according to AP News.
Shares of Palantir jumped more than 20% Tuesday following a strong earnings report — complete with a rosy revenue outlook — on Monday, according to CNBC.

VP, Life Science Cloud Strategy: Salesforce, Philadelphia, PA
Business Development Manager, AI Partnerships: Meta, New York, NY

Source: Unsplash

The major U.S. stock indices moved higher throughout Election Day on Tuesday, with the S&P 500, Nasdaq and Dow Jones Industrial Average all closing the day up more than 1%.

The push was largely led by large-cap tech stocks, with Nvidia finishing the session up 2.8% to overtake Apple as the world’s largest company by market cap, Microsoft up .7%, Meta up 2.1%, Tesla up 3.5% and Apple up .65%.

Over the last three months, the S&P is up more than 10% and the Nasdaq is up more than 12%.

In the midst of this stock reaction, Wedbush analyst Dan Ives said in a note that both Big Tech and Wall Street are watching the election results with “white knuckles” owing to the radically different policy dynamics ahead.

Ives said that investors around the world are worried about the implications of former President Donald Trump re-taking the White House, due to Trump’s proposed tariffs and harsher stance on China, which could “significantly impact the supply chain … and slow the pace of the AI Revolution.”
The other element of this has to do with regulation. Though AI policy has yet to become a major political issue, it is growing in prominence even as the federal government remains stalled out on how to approach regulating the technology.

Ives said that, from an AI perspective, a win from Vice President Kamala Harris would be “more bullish (on) margins” while a Trump win would be a “net negative” for Big Tech.

However, he added that a political gridlock, resulting from the many seats in the Senate and House that are up for grabs, could “complicate any major policy changes for Big Tech in the near term. A red sweep could be the most bullish for the overall market, but would have underlying negative consequences for the Big Tech trade in our view.”

When it comes to AI, Trump has promised to repeal President Joe Biden’s executive order on AI, while Harris is interested in balancing innovation with risk mitigation.

Source: Created with AI by The Deep View

When discussing artificial intelligence in the context of the legal profession, it often concerns some mix of chatbots designed to bring (likely faulty) legal advice to folks at a far more accessible cost and language models designed to help lawyers do their jobs.

The problem with both of these scenarios is the same: the generative AI that we know today is known to output hallucinatory, biased fictions, making it a potentially dangerous tool when dealing with something as high-stakes as the law.

So, when Dan Rabinowitz was founding the legal tech firm Pre/Dicta, he decided to focus on a tried-and-true method: behavioral analytics.

The problem at hand here is one of predictions and predictability: often, clients will ask their legal counsel to predict the outcome of a given motion or trial. Even if they don’t ask, it’s an important thing for lawyers to be able to do, since the odds of a certain outcome will necessarily influence or adjust strategic approaches.

But the reality, Rabinowitz, a former trial attorney for the U.S. Department of Justice and CEO of Pre/Dicta, told me, is that lawyerly predictions aren’t necessarily reliable. Predictive models, however, are more so.

Predictive algorithms and behavioral analytics represent the first heavily scaled and consumer-accepted application of artificial intelligence.
For years, everything from Netflix to social media to Google search has used these algorithms to tailor feeds and home pages and personalize advertising. It’s a relatively basic concept, wherein personal data and internet habits are parsed together to make cold, statistical predictions about future behavior.

Taking this concept — and that base technology — Rabinowitz set out to create predictive models for judge behavior; parsing a number of data inputs, Pre/Dicta’s system can identify patterns in past decisions to make predictions — with an 85% accuracy rate — concerning future outcomes across the lifecycle of a given litigation.

“Our AI is designed — and I say this and sometimes people are disappointed — neither to be more efficient, nor to replace,” Rabinowitz said. “It's not going to save you money because we do what you don't do anyways, and it's not going to replace you because you still should be focused on being a lawyer. But guessing what a judge should be doing, you should never be doing that, ever. That’s what we do.”

For Rabinowitz, the key here is trust. The tech on display — whose predictions are independently checked and verified — isn’t the same as generative AI. It’s been in play for years; it doesn’t pose the same risks.

“What you're really looking for is trustworthy insights. It has to be trustworthy,” he said. “In order to deploy technology, it's not enough simply to take whatever the latest and greatest is … it doesn't necessarily mean it's commercialized at that point in time. It doesn't mean that you should be deploying that into a field that requires a high level of sophistication, significant subject matter expertise, and just go ahead and launch that because you're in an arms race to say that you're the first.”