Wow - the model really hallucinates without hesitation. I asked a number of "What do you know about [person, company, etc]?" questions and rather than realizing that it didn't know about them, it just made up an answer for every one of them.
Come build systems (TypeScript, Postgres, ElasticSearch, Redis, AWS, Redshift, etc) that improve the lives of the neediest patients in the U.S. healthcare system through the power of behavioral economics.
I wonder to what degree an LLM could now produce frames/slots/values in the knowledge graph. With so much structure already existing in the Cyc knowledge graph, could those frames act as the crystal seed upon which an LLM could crystallize its latent knowledge about the world from the trillions of tokens it was trained upon?
I had the same thought. Does anybody know if there have been attempts either to incorporate Cyc-like graphs into LLM training data or to extend such graphs with LLMs?
From time to time, I read articles on the boundary between neural nets and knowledge graphs like a recent [1]. Sadly, no mention of Cyc.
I'd bet, judging mostly from my failed attempts at playing with OpenCyc around 2009, is that the Cyc has always been too closed and to complex to tinker with. That doesn't play nicely with academic work. When people finish their PhDs and start working for OpenAI, they simply don't have Cyc in their toolbox.
The problem is not one of KB size. The Cyc KB is huge. The problem is that the underlying inferencing algorithms don't scale whereas the transformer algorithm does.
Wellth | Mobile Engineer | Remote (US), Los Angeles HQ | Full-time | https://wellthapp.com/
Technologies: React Native, TypeScript, GraphQL
Our smartphone app uses behavioral economics to help sick patients get better - so that health plans can save millions of dollars per year and patients can get paid. We need you to help us add awesome new features to our mobile app!
We raised our Series B a few months ago and are gradually growing our team as we ~double revenue each year. We hire for the long-term; many of our employees have been with us for more than four years. Come work with us and do well by doing good!
Wellth | Playa Vista, CA | Director of Engineering
50% of chronic disease patients don't take their medications as prescribed. We motivate that 50% with a tech-enabled rewards program, delivered through a mobile app and image recognition back-end, by paying each person up to $30/month. That gets them back on track, healthy, and keeps them out of the hospital.
We just raised a $7mm round and are positioning ourselves for the next two years of growth. We need you to come lead our engineering team to success.
The reality is that you need to be able to trust the founders and board members you work with. A corporation's board typically has the authority to issue new shares (thus diluting existing shareholders), so even having unrestricted common stock or even preferred stock does not guarantee that you will always own a certain % of a company's outstanding shares.
Between all the preferred share terms, warrants, options, contingent earn-outs, etc, there's a million ways to engineer the divvying-up of proceeds in a sale.
Ultimately, it comes down to whether you trust your leaders to hold to the spirit of their agreement with you.
All that said... it would be really interesting to see a corporation with bylaws and governance/voting structured in such a way that employee ownership and payout %s could be maintained within certain guaranteed bands. It would severely limit that corporation's flexibility, and might lead it to have a higher risk of failure in bad funding market environments... but it could become a big draw for high-value employees who strongly prefer equity certainty over cash comp and company autonomy.
This is 100% true. You need mechanisms that validate and verify that trust (real stock or standard options, documented legal contracts, etc..) AND you need to have the personal trust in the board/founders in the first place.
Even with both there's risk beyond just market risk, but at least you've done what you can to minimize it.
Wellth is a 4.5yo healthcare tech startup that builds mobile applications to remind and motivate chronic disease patients to take their medications on time and use their medical devices, which keeps them out of the hospital and avoids billions of dollars of unnecessary spending. Our customers are health insurers, risk-bearing hospital systems, and life insurers. We've raised $7.1mm to date from a mix of long-term-view healthcare and life insurance investors (most recently $5mm in Sep '18), and have had a product live in the market for over two years.
We build mobile apps in React Native, front-ends in React, use Apollo and GraphQL, a Postgres DB, and Docker images for deployment onto AWS. We're an agile product/engineering team of 7, have a dedicated head of product, run two-week sprints, and have CI/CD across the stack.
We're looking for a senior engineer with strong back-end experience and at least moderate front-end experience. Bonus points for experience in healthcare or other regulated industries, and for interest in mentoring other engineers and improving engineering processes. 2019 is the year where we prepare for significant scale; help us build kick-ass systems that will live for a long time.
Wellth (https://wellthapp.com) | Los Angeles, CA | Lead Engineer | Full-time | Onsite
I'm the co-founder and CTO of Wellth, a fast-growing, venture-backed healthcare technology startup that is solving the multibillion dollar problem of medication non-adherence. We help patients stick to their care plan and have better health outcomes, saving their hospitals and insurers millions of dollars each year - and ultimately saving lives as a direct result of our work.
We're a team of ten, half of that product engineers and designers, and we're looking for an experienced Lead Engineer to own our technical architecture, build awesome new features, and mentor our talented development team.
Our tech stack is cutting-edge JS: React Native, React, Node.js, GraphQL, Postgres, with some TypeScript and Flow for good measure.
I still think the defining moment for ML inference (and maybe even training!) on embedded devices will come when there are viable special-purpose, low-power ML chips.
As much as I hate to do this, I'm going to make a comparison to Bitcoin mining.
Mining is all about optimizing hashes/joule to get the best ROI. We watched it go from CPU -> GPU -> FPGA -> ASIC in the quest for efficiency.
But I think the final leap will come by going from digital execution to application-specific analog computing. If you don't need high precision, you can compute extremely quickly and efficiently using properly-configured analog circuits.
I remain unconvinced we'll see ASICs dominating inference. Part of the problem is that even if we're just talking about neural networks, there's a variety of architectures, activation functions, etc. to consider. At this stage, from my own benchmarking Nvidia is close enough to the TPU with the V100 card while allowing much more flexibility in the software stack used.
For inference, GPUs are also pretty damn efficient since it's an embarrassingly parallel task w/ minimal synchronization (no gradient updates needed). In this case, FPGAs are a far better choice since you can push updates to accommodate new network architectures, activation functions, ,etc. The TPU instead relies on a matrix-multiplier unit which supports more use cases but won't be as performant on something like an RNN.
After some investigation, you are correct! Knowing that some of TrueNorth's creators previously worked on mixed-mode systems, I made the assumption that this one was too.
It seems the TrueNorth is indeed fully digital, but takes advantage of the event-driven architecture and peer-to-peer communication between many tiny cores to keep things low-power.
Wellth | Lead Engineer | Venice, CA | Onsite / Remote | Full-time
Wellth (https://wellthapp.com/) helps patients stay healthy after they leave the hospital by rewarding them for taking their medications and taking good care of themselves. We're funded by large insurers and healthcare VCs, and our mobile apps have shown great results in clinical populations with diabetes, heart failure, or heart attack discharges. We love hearing from our users how the app has changed their lives for the better!
We're based in Brooklyn currently, but moving part of the team to the Los Angeles area (probably Venice or Santa Monica) in a few months. We're looking for a great Lead Engineer who can contribute to all areas of our stack, mentor junior developers, and help whip our engineering processes into even better shape in tandem with our CTO. Looking to hire anytime in the next few months, with the possibility of starting remote until we move to LA.
We're React Native for our mobile apps (iOS and Android), React for our admin dashboard, with Apollo and GraphQL for both, connected to a Node.js backend with Postgres / Postgraphile, deployed to a HIPAA-compliant PaaS. The current product team is two mobile engineers, one web engineer, a designer, and the CTO who does back-end and AI work.
Come help us build awesome, engaging applications that help patients stay healthy and lower the healthcare cost burden in the US!
