IFTF's Post Financial Futures Forecasts using GPT-3

Worked for years with IFTF (Institute for the Future) in the enterprise.

The Institute for the Future passes along this new GPT-3 Forecast Model, a nice use example.   I will pass along additional findings.

From IFTF: 

We also conducted a groundbreaking experiment using our forecast provocation and OpenAI's GPT-3 to generate over 1,400 live, unique scenarios of post-financialization futures. If you missed it or would like to try out a scenario again the link is here.  

This is an experimental prototype exploring the use of GPT-3 to help augment and customize IFTF’s human forecasting process to broader audiences and more interactive contexts. 

IFTF has pursued this experiment in the spirit of building our literacy of the capabilities and challenges of large language models, and this experiment should not be construed as an endorsement of OpenAI or GPT-3.

IFTF is not responsible for content produced by the large language model GPT-3. The content provided by GPT-3 does not reflect the opinions or views of IFTF.  (Nor of Franz Dill) 

Be sure to mark your calendars for the next IFTF Ten-Year Forecast series event coming up on June 1. We'll be announcing details on expert speakers and special interactive futures immersions very soon.  

Thank you for being a part of IFTF’s TYF community as we look toward shaping a better future!

Our mailing address is:  Institute for the Future,  201 Hamilton Avenue,   Palo Alto, CA 94301  Iftf.org

DeepMind & Google’s ML-Based GraphCast

Did a quick updated look,  hard to get a useful error track of forevast.

DeepMind & Google’s ML-Based GraphCast Outperforms the World’s Best Medium-Range Weather Forecasting System  in Medium.com

Medium-range weather forecasts play a crucial role in agriculture, construction, travel and other industries. They also bring practical value to people’s daily lives, enabling us to plan outings and keeping us safe from extreme weather events. Traditional numerical weather prediction (NWP)-based forecasting models that run simulations on computing clusters however do not scale efficiently with today’s increasing weather data availability, and their accuracy relies on manual input from experts, which is time-consuming and cost inefficient.

In the new paper GraphCast: Learning Skillful Medium-Range Global Weather Forecasting, a research team from DeepMind and Google presents GraphCast, a machine-learning (ML)-based weather simulator that scales well with data and can generate a 10-day forecast in under 60 seconds. GraphCast outperforms the world’s most accurate deterministic operational medium-range weather forecasting system and all existing ML-based benchmarks. ... '

Deciding When to Trust

Have recently been involved with the concept of  'smart contracts' and trustble agreements.    How might this idea be included in the construction of such things?  Things that we can trust in a neuroscience sense?  Is trust just a way to accurately forecast an agent's future behavior?

How do Our Brains Decide When to Trust?      By Paul J. Sak, in the HBR

Trust is the enabler of global business — without it, most market transactions would be impossible. It is also a hallmark of high-performing organizations. Employees in high-trust companies are more productive, are more satisfied with their jobs, put in greater discretionary effort, are less likely to search for new jobs, and even are healthier than those working in low-trust companies. Businesses that build trust among their customers are rewarded with greater loyalty and higher sales. And negotiators who build trust with each other are more likely to find value-creating deals.

Despite the primacy of trust in commerce, its neurobiological underpinnings were not well understood until recently. Over the past 20 years, research has revealed why we trust strangers, which leadership behaviors lead to the breakdown of trust, and how insights from neuroscience can help colleagues build trust with each other — and help boost a company’s bottom line.

THE BIOLOGY OF TRUST

Human brains have two neurological idiosyncrasies that allow us to trust and collaborate with people outside our immediate social group (something no other animal is capable of doing). The first involves our hypertrophied cortex, the brain’s outer surface, where insight, planning, and abstract thought largely occur. Parts of the cortex let us do an amazing trick: transport ourselves into someone else’s mind. Called theory of mind by psychologists, it’s essentially our ability to think, “If I were her, I would do this.” It lets us forecast others’ actions so that we can coordinate our behavior with theirs.  ..... "   (Details at the ink)

Explaining Simplistic Forecasts to Management

Recall this kind of thing coming up many times.   We addressed it by installing a clickable pop up to anticipate the question and make the case for the forecast.    Agree, its about selling the forecast,in context to management.  Good thoughts on the issue below from SAS:

How do I explain a flat-line forecast to senior management?   By Charlie Chase   SAS

How do you explain flat-line forecasts to senior management? Or, do you just make manual overrides to adjust the forecast?   

When there is no detectable trend or seasonality associated with your demand history, or something has disrupted the trend and/or seasonality, simple time series methods (i.e. naïve and simple exponential smoothing) will often generate a flat-line forecast reflecting the current demand level. Because a flat-line is often an unlikely reflection of the future, delivering a flat-line forecast to management may require explanation. And sometimes, explaining is not enough.

Today, we have large scale automatic hierarchical statistical forecasting systems to automatically build statistical models up/down a business hierarchy for hundreds of thousands, and in some cases millions, of data series. As you add more historical data, and causal factors (i.e., price, promotions, advertising, in-store merchandizing, economic data and others), the system re-diagnoses this information and rebuilds (tweaks) the models automatically. They also automatically identify and correct for outliers and other anomalies in the demand history.

The ability to use stacked neural network (NN) plus time series models have proven to be the best forecasting method according to the recent M4 competition. Stacked NN + time series ensemble models are just another statistical method that can be used along with traditional methods (e.g. naïve, exponential smoothing, ARIMA, ARIMAX, dynamic regression, unobserved components models, weighted combined models and others).

We all know how hard it is to beat a naïve model over time. As a result, naïve models are now the benchmark for evaluating forecasts. If your forecast can’t beat a naïve model, then why are you spending so much time developing and adjusting (manual overrides) statistical forecasts?
Subsequently, we all know that not all products are forecastable using statistical methods because of sparse data, randomness, lack of historical demand data, and no access to causal information. However, it’s not just a matter of forecast accuracy, but also whether you can sell the forecast to senior management. ..... "

State of AI Report

Looks to be a good resource,  subscribe.

State of AI Report 2019  By Nathan Benaich and Ian Hogarth
Artificial intelligence (AI) is a multidisciplinary field of science whose goal is to create intelligent machines.

We believe that AI will be a force multiplier on technological progress in our increasingly digital, data-driven world. This is because everything around us today, ranging from culture to consumer products, is a product of intelligence.

In this report, we set out to capture a snapshot of the exponential progress in AI with a focus on developments in the past 12 months. Consider this report as a compilation of the most interesting things we’ve seen that seeks to trigger an informed conversation about the state of AI and its implication for the future. This edition builds on the inaugural State of AI Report 2018, which can be found here.

We consider the following key dimensions in our report:

Research: Technology breakthroughs and their capabilities.
Talent: Supply, demand and concentration of talent working in the field.
Industry: Large platforms, financings and areas of application for AI-driven innovation today and tomorrow.
China: Large platforms, financings and areas of application for AI-driven innovation today and tomorrow.
Politics: Public opinion of AI, economic implications and the emerging geopolitics of AI.

Read and download the State of AI Report 2019 and 2018 on SlideShare.
Collaboratively produced in East London, UK by:

Nathan Benaich (@nathanbenaich)    Ian Hogarth (@soundboy)

Top Consumer Goods Companies

Top 100 Consumer Goods Companies 2017
By CGT Staff - 12/27/2017

How tough is the consumer goods market these days? Almost one-third of the world’s largest public consumer goods companies experienced sales declines in 2016. Another 25 leading companies posted growth of 2.0% or less.

That means the annual letter to shareholders didn’t deliver much in the way of good financial news at more than half of the companies listed in CGT’s "Top 100 Consumer Goods Companies" for 2017.
These decidedly disappointing results confirm the research at IDC Manufacturing Insights, which finds that only 3.0% of the $35 billion in net growth in the consumer goods industry over the last three years has come from traditional, large enterprise players.  .... "


I have worked directly for or with projects for several of the top 5.  - Franz

Machine Learning is Wrong

An important point.  Like I said in teaching forecasting classes, the forecast is always wrong, and that error should be considered in risk analysis.  It does not mean it is not useful.  Just always incorrect in any significant application.  ML is also not magic. Good discussion below.

The machine learning paradox
Nothing says machine learning can't outperform humans, but it's important to realize perfect machine learning doesn't, and won't, exist.     By Mike Loukides June 1, 2017 .... 

AI, Robots Won't Try To Kill Us

I mentioned this Stanford study when it came out.   Worth the read.  Now Fast Company takes a closer look.   'Try' is the operative word. Long view is right, with some interesting but hard to calibrate forecasts.

Robots Won't Try To Kill Us, Says Stanford's 100-Year Study Of AI
Stanford University is taking the long view on the future of AI with an ambitious project that will unfold throughout the century.  .... " 

Using Twitch to Predict the Future

Can Twitch teach us the future?    New to Twitch but unclear it can do this consistently.  Any data contains elements of the future within it.   What if you observed any kind of activity, not just gaming, what does that teach us?   I still need something measurable.

" ... If you devote 20 solid hours to predictions over the next 6 months; I’m 70% confident you’ll evaluate yourself as being 10% better at predicting the future when you review your predictions.
Think about what 10% more foresight is worth to you, wherever you need it most. ... " 

The Future of AI in the Home, Office and City

An area we covered for years.  How AI will influence will our home, business and urban worlds by 2030?

Stanford-Hosted Study Examines How AI Might Affect Urban Life in 2030 Stanford News (09/01/16) Tom Abate 

A year-long, Stanford University-hosted study projects how artificial intelligence (AI) will realistically impact North American urban life in eight domains by 2030. Stanford's One Hundred Year Study on AI (AI100) is the result of a standing committee of researchers enlisted to evaluate the technological, economic, and policy ramifications of potential AI applications in a societally relevant environment. Five sections of the new report focus on application areas such as transportation, home/service robots, healthcare, education, and entertainment. The three other sections concentrate on technological effects in domains such as low-resource communities, public safety and security, and employment and the workplace. The AI100 panel says the study seeks to guide ethical development of AI technologies via public discourse. "It is not too soon for social debate on how the fruits of an AI-dominated economy should be shared," the report's authors note. AI100 standing committee chair Barbara Grosz, recipient of the 2009 ACM AAAI Allen Newell Award, says AI technologies can be reliable and yield a wide spectrum of benefits. "Being transparent about their design and deployment challenges will build trust and avert unjustified fear and suspicion," she notes. .... " 

Train Your Brain to Think About the Future

We all do this, but how well?  Here looking at it early.

Train Your Brain to Think About the Future by Katie King

What fraction of your time is spent thinking about the future, and how far out do you think? Last month, this question kicked off a lively two-day workshop with K-12 leaders about our newest forecast,  .... 

" ... In the meantime, the discussion that arose from that question stuck with me. Most of our participants said they think no longer than 1-5 years ahead, and most spend less than half of their time thinking beyond the present. There’s nothing wrong with that; nothing would get done if we all sat around thinking about what might happen in 25 years. That said, those of us who care about education aim to prepare students for the future, so we should know how to think about it and do so with some regularity. ..... " .  

Fedex Holiday Glitch

Over holiday,  Fedex apparently unable to meet increased demand.  Poor forecasting of demand, resource allocation?   Awaiting more information.  Might be an interesting case study for highly varying demand.

Watson Trends looks at Holiday Toy Shopping

In Forbes.  Looks fairly obvious,  but it is  the specific forecast number that make a difference to the elements of the supply chain involved.

Watson Based Trend App

In Datanami: An app for the holidays that tracks consumer trends.

" ... The app seeks to create another consumer use case for the Watson cognitive technology platform that leverages natural language processing and machine learning to glean insights from heaps of unstructured data. In the case of Watson Trend, IBM hopes to offer consumers sounder advice on, say, what color tie to buy hubby, than traditional social media pointers designed to advise consumers on purchasing decisions.

“With this app we are trying out the idea that Watson’s insights can go beyond the usual gift guide to help users find and give more inspired gifts,” IBM declared.  " ... '  

Watson Trend App:
" ... Put a personal shopper in your pocket
At IBM, we wanted to see if putting the cognitive power of Watson in the hands of consumers could make their holiday season easier. Use the app to go beyond best seller gift guides. Get daily feeds about what products are trending and why. See forecasts for where trends are headed. And share with us your feedback so we can make Watson the most valuable personal shopper to ever fit in your pocket.  ... " 

Digital to Overtake TV Ad Spending in 2 Years

In Adage:  Comparing  digital to TV ads, a prediction.  But how about streaming TV via digital?

Executives Understanding Emerging Technology

I spent years explaining emerging technology to the C-Suite.  So this article in CIO Insight reminded me of the complexity involved. We also worked this during the emergence of the commercial Internet, which started with executive disbelief,  led to unrealistic expectations, then had the technology dominating both internal and external operations.  Who could have predicted the extent and speed of these changes?  We are still in the midst of this, so doing a good job of technology forecast remains  important.

Pipelines and Forecast Timing

In E-Commercetimes: The timing of forecast value is an important aspect.  It's part of the decision process that is often forgotten.

" ... Analytics should be able to tell you the likelihood of each deal's completion in time for you to take action if the prediction is unfavorable. If each deal is somewhat different, then comparing each with the historic deals most like them will provide a truer indicator of close probability than simply comparing deals through something as general-purpose as deal stage. The difference is important. ... "