AI needs the constraints of the human brain

Danyal Akarca

iai.tv

Originally posted 30 Jan 24

Here is an excerpt:

So, evolution shapes systems that are capable of solving competing problems that are both internal (e.g., how to expend energy) and external (e.g., how to act to survive), but in a way that can be highly efficient, in many cases elegant, and often surprising. But how does this evolutionary story of biological intelligence contrast with the current paradigm of AI?

In some ways, quite directly. Since the 50s, neural networks were developed as models that were inspired directly from neurons in the brain and the strength of their connections, in addition to many successful architectures of the past being directly motivated by neuroscience experimentation and theory. Yet, AI research in the modern era has occurred with a significant absence of thought of intelligent systems in nature and their guiding principles. Why is this? There are many reasons. But one is that the exponential growth of computing capabilities, enabled by increases of transistors on integrated circuits (observed since the 1950s, known as Moore’s Law), has permitted AI researchers to leverage significant improvements in performance without necessarily requiring extraordinarily elegant solutions. This is not to say that modern AI algorithms are not widely impressive – they are. It is just that the majority of the heavy lifting has come from advances in computing power rather than their engineered design. Consequently, there has been relatively little recent need or interest from AI experts to look to the brain for inspiration.

But the tide is turning. From a hardware perspective, Moore’s law will not continue ad infinitum (at 7 nanometers, transistor channel lengths are now nearing fundamental limits of atomic spacing). We will therefore not be able to leverage ever improving performance delivered by increasingly compact microprocessors. It is likely therefore that we will require entirely new computing paradigms, some of which may be inspired by the types of computations we observe in the brain (the most notable being neuromorphic computing). From a software and AI perspective, it is becoming increasingly clear that – in part due to the reliance on increases to computational power – the AI research field will need to refresh its conceptions as to what makes systems intelligent at all. For example, this will require much more sophisticated benchmarks of what it means to perform at human or super-human performance. In sum, the field will need to form a much richer view of the possible space of intelligent systems, and how artificial models can occupy different places in that space.

The info is here.

Key Points:

• Evolutionary pressures: Efficient, resource-saving brains are advantageous for survival, leading to optimized solutions for learning, memory, and decision-making.
• AI's reliance on brute force: Modern AI often achieves performance through raw computing power, neglecting principles like energy efficiency.
• Shifting AI paradigm: Moore's Law's end and limitations in conventional AI call for exploration of new paradigms, potentially inspired by the brain.
• Neurobiology's potential: Brain principles like network structure, local learning, and energy trade-offs can inform AI design for efficiency and novel functionality.
• Embodied AI with constraints: Recent research incorporates space and communication limitations into AI models, leading to features resembling real brains and potentially more efficient information processing.

Health Care in the U.S. Compared to Other High-Income Countries: Worst Outcomes

The Commonwealth Fund

Mirror, Mirror 2021: Reflecting Poorly

Originally posted 4 Aug 21

Introduction

No two nations are alike when it comes to health care. Over time, each country has settled on a unique mix of policies, service delivery systems, and financing models that work within its resource constraints. Even among high-income nations that have the option to spend more on health care, approaches often vary substantially. These choices affect health system performance in terms of access to care, patients’ experiences with health care, and people’s health outcomes. In this report, we compare the health systems of 11 high-income countries as a means to generate insights about the policies and practices that are associated with superior performance.

With the COVID-19 pandemic imposing an unprecedented stress test on the health care and public health systems of all nations, such a comparison is especially germane. Success in controlling and preventing infection and disease has varied greatly. The same is true of countries’ ability to address the challenges that the pandemic has presented to the workforce, operations, and financial stability of the organizations delivering care. And while the comparisons we draw are based on data collected prior to the pandemic or during the earliest months of the crisis, the prepandemic strengths and weaknesses of each country’s preexisting arrangements for health care and public health have undoubtedly been shaping its experience throughout the crisis.

For our assessment of health care system performance in Australia, Canada, France, Germany, the Netherlands, New Zealand, Norway, Sweden, Switzerland, the United Kingdom, and the United States, we used indicators available across five domains:

• Access to care
• Care process
• Administrative efficiency
• Equity
• Health care outcomes

For more information on these performance domains and their component measures, see How We Measured Performance. Most of the data were drawn from surveys examining how members of the public and primary care physicians experience health care in their respective countries. These Commonwealth Fund surveys were conducted by SSRS in collaboration with partner organizations in the 10 other countries. Additional data were drawn from the Organisation for Economic Co-operation and Development (OECD) and the World Health Organization (WHO).

The info is here.

What One Health System Learned About Providing Digital Services in the Pandemic

Marc Harrison

Harvard Business Review

Originally posted 11 Dec 20

Here are two excerpts:

Lesson 2: Digital care is safer during the pandemic.

A patient who’s tested positive for Covid doesn’t have to go see her doctor or go into an urgent care clinic to discuss her symptoms. Doctors and other caregivers who are providing virtual care for hospitalized Covid patients don’t face increased risk of exposure. They also don’t have to put on personal protective equipment, step into the patient’s room, then step outside and take off their PPE. We need those supplies, and telehealth helps us preserve it.

Intermountain Healthcare’s virtual hospital is especially well-suited for Covid patients. It works like this: In a regular hospital, you come into the ER, and we check you out and think you’re probably going to be okay, but you’re sick enough that we want to monitor you. So, we admit you.

With our virtual hospital — which uses a combination of telemedicine, home health, and remote patient monitoring — we send you home with a technology kit that allows us to check how you’re doing. You’ll be cared for by a virtual team, including a hospitalist who monitors your vital signs around the clock and home health nurses who do routine rounding. That’s working really well: Our clinical outcomes are excellent, our satisfaction scores are through the roof, and it’s less expensive. Plus, it frees up the hospital beds and staff we need to treat our sickest Covid patients.

(cut)

Lesson 4: Digital tools support the direction health care is headed.

Telehealth supports value-based care, in which hospitals and other care providers are paid based on the health outcomes of their patients, not on the amount of care they provide. The result is a greater emphasis on preventive care — which reduces unsustainable health care costs.

Intermountain serves a large population of at-risk, pre-paid consumers, and the more they use telehealth, the easier it is for them to stay healthy — which reduces costs for them and for us. The pandemic has forced payment systems, including the government’s, to keep up by expanding reimbursements for telehealth services.

This is worth emphasizing: If we can deliver care in lower-cost settings, we can reduce the cost of care. Some examples:

• The average cost of a virtual encounter at Intermountain is $367 less than the cost of a visit to an urgent care clinic, physician’s office, or emergency department (ED).
• Our virtual newborn ICU has helped us reduce the number of transports to our large hospitals by 65 a year since 2015. Not counting the clinical and personal benefits, that’s saved $350,000 per year in transportation costs.
• Our internal study of 150 patients in one rural Utah town showed each patient saved an average of $2,000 in driving expenses and lost wages over a year’s time because he or she was able to receive telehealth care close to home. We also avoided pumping 106,460 kilograms of CO2 into the environment — and (per the following point) the town’s 24-bed hospital earned $1.6 million that otherwise would have shifted to a larger hospital in a bigger town.

The info is here.

Ethics In AI: Why Values For Data Matter

Marc Teerlink
forbes.com
Originally posted 18 Dec 19

Here is an excerpt:

Data Is an Asset, and It Must Have Values

Already, 22% of U.S. companies have attributed part of their profits to AI and advanced cases of (AI infused) predictive analytics.

According to a recent study SAP conducted in conjunction with the Economist’s Intelligent Unit, organizations doing the most with machine learning have experienced 43% more growth on average versus those who aren’t using AI and ML at all — or not using AI well.

One of their secrets: They treat data as an asset. The same way organizations treat inventory, fleet, and manufacturing assets.

They start with clear data governance with executive ownership and accountability (for a concrete example of how this looks, here are some principles and governance models that we at SAP apply in our daily work).

So, do treat data as an asset, because, no matter how powerful the algorithm, poor training data will limit the effectiveness of Artificial Intelligence and Predictive Analytics.

The info is here.

Death by a Thousand Clicks: Where Electronic Health Records Went Wrong

Erika Fry and Fred Schulte
Fortune.com
Originally posted on March 18, 2019

Here is an excerpt:

Damning evidence came from a whistleblower claim filed in 2011 against the company. Brendan Delaney, a British cop turned EHR expert, was hired in 2010 by New York City to work on the eCW implementation at Rikers Island, a jail complex that then had more than 100,000 inmates. But soon after he was hired, Delaney noticed scores of troubling problems with the system, which became the basis for his lawsuit. The patient medication lists weren’t reliable; prescribed drugs would not show up, while discontinued drugs would appear as current, according to the complaint. The EHR would sometimes display one patient’s medication profile accompanied by the physician’s note for a different patient, making it easy to misdiagnose or prescribe a drug to the wrong individual. Prescriptions, some 30,000 of them in 2010, lacked proper start and stop dates, introducing the opportunity for under- or overmedication. The eCW system did not reliably track lab results, concluded Delaney, who tallied 1,884 tests for which they had never gotten outcomes.

(cut)

Electronic health records were supposed to do a lot: make medicine safer, bring higher-quality care, empower patients, and yes, even save money. Boosters heralded an age when researchers could harness the big data within to reveal the most effective treatments for disease and sharply reduce medical errors. Patients, in turn, would have truly portable health records, being able to share their medical histories in a flash with doctors and hospitals anywhere in the country—essential when life-and-death decisions are being made in the ER.

But 10 years after President Barack Obama signed a law to accelerate the digitization of medical records—with the federal government, so far, sinking $36 billion into the effort—America has little to show for its investment.

The info is here.

Optimality bias in moral judgment

Julian De Freitas and Samuel G. B. Johnson
Journal of Experimental Social Psychology
Volume 79, November 2018, Pages 149-163

Abstract

We often make decisions with incomplete knowledge of their consequences. Might people nonetheless expect others to make optimal choices, despite this ignorance? Here, we show that people are sensitive to moral optimality: that people hold moral agents accountable depending on whether they make optimal choices, even when there is no way that the agent could know which choice was optimal. This result held up whether the outcome was positive, negative, inevitable, or unknown, and across within-subjects and between-subjects designs. Participants consistently distinguished between optimal and suboptimal choices, but not between suboptimal choices of varying quality — a signature pattern of the Efficiency Principle found in other areas of cognition. A mediation analysis revealed that the optimality effect occurs because people find suboptimal choices more difficult to explain and assign harsher blame accordingly, while moderation analyses found that the effect does not depend on tacit inferences about the agent's knowledge or negligence. We argue that this moral optimality bias operates largely out of awareness, reflects broader tendencies in how humans understand one another's behavior, and has real-world implications.

The research is here.

How AI is transforming the NHS

Ian Sample
The Guardian
Originally posted July 4, 2018

Here is an excerpt:

With artificial intelligence (AI), the painstaking task can be completed in minutes. For the past six months, Jena has used a Microsoft system called InnerEye to mark up scans automatically for prostate cancer patients. Men make up a third of the 2,500 cancer patients his department treats every year. When a scan is done, the images are anonymised, encrypted and sent to the InnerEye program. It outlines the prostate on each image, creates a 3D model, and sends the information back. For prostate cancer, the entire organ is irradiated.

The software learned how to mark up organs and tumours by training on scores of images from past patients that had been seen by experienced consultants. It already saves time for prostate cancer treatment. Brain tumours are next on the list.

Automating the process does more than save time. Because InnerEye trains on images marked up by leading experts, it should perform as well as a top consultant every time. The upshot is that treatment is delivered faster and more precisely. “We know that how well we do the contouring has an impact on the quality of the treatment,” Jena says. “The difference between good and less good treatment is how well we hit the tumour and how well we avoid the healthy tissues.”

The article is here.

Is it dangerous for humans to depend on computers?

Rory Cellan-Jones
BBC News
Originally published June 5, 2017

Here is an excerpt:

In Britain, doctors whose computers froze during the recent ransomware attack had to turn patients away. In Ukraine, there were power cuts when hackers attacked the electricity system, and five years ago, millions of Royal Bank of Scotland customers were unable to get at their money for days after problems with a software upgrade.

Already some people have had enough. This week a letter to the Guardian newspaper warned that the modern world was "dangerously exposed by this reliance on the internet and new technology".
The correspondent, quite possibly a retired government employee, continued "there are just enough old-time civil servants left alive to turn back the clock and take away our dangerous dependence on modern technology."

Somehow, though, I don't see this happening. Airlines are not going to scrap the computers and tick passengers off on a paper list before they climb aboard, bank clerks will not be entering transactions in giant ledgers in copperplate writing.

In fact, computers will take over more and more functions once restricted to humans, most of them far more useful than a game of Go. And that means that at home, at work and at play we will have to get used to seeing our lives disrupted when those clever machines suffer the occasional nervous breakdown.

The article is here.

The Specialists’ Stranglehold on Medicine

Jamie Koufman
The New York Times - Opinion
Originally posted June 3, 2017

Here is an excerpt:

Neither the Affordable Care Act nor the Republicans’ American Health Care Act addresses the way specialists are corrupting our health care system. What we really need is what I’d call a Health Care Accountability Act.

This law would return primary care to the primary care physician. Every patient should have one trusted doctor who is responsible for his or her overall health. Resources must be allocated to expand those doctors’ education and training. And then we have to pay them more.

There are approximately 860,000 practicing physicians in the United States today, and too few — about a third — deliver primary care. In general, they make less than half as much money as specialists. I advocate a 10 percent to 20 percent reduction in specialist reimbursement, with that money being allocated to primary care doctors.

Those doctors should have to approve specialist referrals — they would be the general contractor in the building metaphor. There is strong evidence that long-term oversight by primary care doctors increases the quality of care and decreases costs.

The bill would mandate the disclosure of procedures’ costs up front. The way it usually works now is that right before a medical procedure, patients are asked to sign multiple documents, including a guarantee that they will pay whatever is not covered by insurance.  But they will have no way of knowing what the procedure actually costs. Their insurance may cover 90 percent, but are they liable for 10 percent of $10,000 or $100,000?

We also need more oversight of those costs. Instead of letting specialists’ lobbyists set costs, payment algorithms should be determined by doctors with no financial stake in the field, or even by non-physicians like economists. An Independent Payment Advisory Board was created by Obamacare; it should be expanded and adequately funded.

The article is here.

The Simple Economics of Machine Intelligence

Ajay Agrawal, Joshua Gans, and Avi Goldfarb
Harvard Business Review
Originally published November 17, 2016

Here are two excerpts:

The first effect of machine intelligence will be to lower the cost of goods and services that rely on prediction. This matters because prediction is an input to a host of activities including transportation, agriculture, healthcare, energy manufacturing, and retail.

When the cost of any input falls so precipitously, there are two other well-established economic implications. First, we will start using prediction to perform tasks where we previously didn’t. Second, the value of other things that complement prediction will rise.

(cut)

As machine intelligence improves, the value of human prediction skills will decrease because machine prediction will provide a cheaper and better substitute for human prediction, just as machines did for arithmetic. However, this does not spell doom for human jobs, as many experts suggest. That’s because the value of human judgment skills will increase. Using the language of economics, judgment is a complement to prediction and therefore when the cost of prediction falls demand for judgment rises. We’ll want more human judgment.

The article is here.