G-O-O-G-L-E Still Rules the Alphabet (BusinessWeek)

One thing hasn’t changed in Google’s metamorphosis into Alphabet. The company’s science projects tickle the imagination, but the same old Google is making all the money. 

Alphabet is the entity created last year to hold everything from Google Web search and Android smartphone software to bets on cars that teach themselves to drive, homes that heat themselves and treatments to help people live forever.

These futuristic projects light up the same “Oh, cool!” part of ours brains that Jules Verne novels did. Just for a couple of minutes, though, tune out Jules Verne.

For now, the only thing that matters for Alphabet is Google and the health of its advertising business. And the real Google, we learned Monday, is more profitable than most people expected. (Although no one really knew or knows anything about the company’s finances. More on that in a moment.) 

Separated from the financial drain of Alphabet's futuristic experiments, Google -- that is, the search engine,  YouTube, Android, etc. -- posted an operating profit margin of 32 percent, before subtracting payments Google makes to its ad partners. That is a level of profitability Google hasn't seen since 2012 under the old financial method that lumped together the money-making and speculative parts of the business. 

Click Power

More people are clicking on Google's Web ads, but the average cost of each click is falling. That's the result of fast-growing sales of ads on smartphones and YouTube, which have lower prices than those on computers.

In the run-up to Alphabet's financial unveiling Monday, it was remarkable how little the world knew about how much the company was splurging on the wannabe businesses like the Nest thermostat, Internet service provider Google Fiber and hot air balloons that ferry Internet service to remote corners of the globe. Wall Street estimates had been all over the map, from operating losses for the "other bets" companies of $500 million to $5 billion. You might as well type “I don't know” into a spreadsheet. It turns out the operating loss last year was $3.6 billion on relatively tiny revenue of $448 million. 

Invisible 'Bets'

Google -- including the Web search engine, YouTube and Android -- is a revenue and profit juggernaut. Alphabet's science projects have a sliver of Google's revenue and $3.6 billion in operating losses.

The disclosure that the "bets" parts of Google are draining billions of dollar a year is oddly good news. That means the core parts of Google are the same as they ever were, and then some: a ruthlessly effective seller of advertising that would make William Randolph Hearst supremely envious, helped along by perhaps the most advanced computerized systems on the planet.

Proving its ad business was even more impressive than meets the eye was exactly the goal of the new Alphabet structure. The company now looks set to pass Apple on Tuesday as the biggest company in the world by stock market value. 

But even for essential parts of the company, we’re still seeing Google through financial Vaseline. What does YouTube’s profit and loss statement look like? Android’s? We didn't know, and we still don't.

Yet the science projects still matter because ambition is the essential fuel of the technology business. Facebook, Apple and Amazon are no slouches in the Jules Verne department, but no other company has the same scope of ambition as Google, which has its fingers in digital media, health care, transportation, Internet connectivity, artificial intelligence, energy and much more.

Companies are obsessed with having a second act, and few can. IBM, Polaroid and Nokia couldn’t be world beaters forever. Google is betting that somewhere in the Alphabet sprawl technologies will emerge that can change the world and sustain Google for the next generations. Some of them are likely to be spectacular flops. Google’s conversion into Alphabet is a real life test of whether a new corporate structure can sustain the audacious imaginations in Mountain View. 

Stock investors, of course, may choose to view “audacious” as code for spendthrift and unfocused. As the company’s costs crept up to hire engineers for wannabe businesses, provide employees with free sushi and buy land for computing hubs, Google last year hired a Wall Street number cruncher to keep a closer tab on the budgets -- or at least to create the impression of financial discipline. In a sign of more careful spending, Google's 2015 tab for real estate, computer server farms and other capital expenditures fell nearly 10 percent in 2015, after tripling from 2011 to 2014.

And indeed that is the central question about Google: Will it wisely use its incredible money-generating advertising machine to find its second acts? The answer is out of reach. Google has always been a “trust us" company. Trust us not to do creepy things with all the information we collect with every click. And trust us that the $3.6 billion profit drain from science projects are a smart bet with shareholder money.

It's clear the Jules Verne projects will not pay off until the distant future, if they ever do. But there's enough in the here-and-now Google to keep profitability humming long enough for imagination to catch up. 

Veterans Florida Entrepreneurship Program Now Accepting Applications

Media Release       Contacts: Gerry Goldstein (850) 898-3489 info@FloridaVetBiz.org   February 1, 2016 For Immediate Release Veterans Florida Entrepreneurship Program Now Accepting Applications New Statewide Initiative Offers Accelerated Entrepreneurship Education to Veterans at No Cost FSBDCN State Office (Pensacola, Fla.) - The Veterans Florida Entrepreneurship Program, a new statewide initiative that offers qualified veterans the opportunity to receive an accelerated entrepreneurship education at no cost, is now accepting applications. The pilot program will begin its first phase in March.  The program is available to veterans who meet the following requirements: have been honorably discharged, are National Guardsmen or Reservists, or are active duty within one year of separation; reside in Florida, and demonstrate a strong interest in entrepreneurship.  The program will be delivered over three phases. The first phase consists of an eight-week online class to help participants build their venture on a strong business foundation. The second phase consists of six days of on-site training at one of the five participating institutions of higher education. This phase will be held on three consecutive weekends and all travel expenses will be covered for participants. The final phase of the program consists of six-months of ongoing mentorship with Florida SBDC Network business consultants. "The Veterans Florida Entrepreneurship Program provides veterans an opportunity to gain a high-quality entrepreneurship education at no cost," said Michael Myhre, CEO and Network State Director for the Florida SBDC. "Our goal is to help our state's veterans navigate the path towards entrepreneurship to fulfill their passions and create jobs and wealth for Florida's economy." The Veterans Florida Entrepreneurship Program is funded through a $1 million appropriation by the Florida Legislature. In December, Veterans Florida, a non-profit created by the Florida Legislature to attract and retain veterans, selected the University of West Florida, through its Florida SBDC Network and Military Veterans Resource Center, to administer the program. The University of West Florida, University of North Florida, Florida Gulf Coast University, Florida Atlantic University, and Hillsborough Community College will serve as network partners of the program. In this role, each institution will dedicate existing small business development resources to deliver the program. "Veterans possess many of the skills necessary to be successful entrepreneurs," said Bobby Carbonell, Executive Director of Veterans Florida. "The technical and financial skills provided from this program, along with the motivation, confidence, attention to detail and problem solving skills they bring, will ensure veterans achieve their vision of the American Dream, owning their own business." Space is limited. The deadline to apply is March 15, 2016. For more information, and to apply, please visit www.FloridaVetBiz.org.  About the Florida SBDC Network:  The Florida SBDC Network, the state's principal provider of business assistance, is celebrating its 40th anniversary in 2016. Since its inception, the Florida SBDC Network has nourished a statewide partnership between higher education and economic development to provide emerging and established business owners with management and technical assistance, enabling overall growth, increased profitability, and economic prosperity for the state. In 2014, the initiatives of the Florida SBDC Network resulted in 42,664 jobs created, retained and saved; $5.8 billion in sales growth; $140.2 million in capital accessed; $210.2 million in government contract awards; and 952 new businesses started. A statewide network of over 40 centers, the Florida SBDC is funded in part by the U.S. Small Business Administration, Defense Logistics Agency, State of Florida and other private and public partners, with the University of West Florida serving as the network's designated lead host institution. The Florida SBDC Network is nationally accredited by the Association of SBDCs. For more information, please visit www.FloridaSBDC.org.   About Veterans Florida: Veterans Florida was established by the Florida Legislature in 2014 as part of the Florida GI Bill to serve as an innovative corporation to encourage recently retired or separated military personnel to make Florida their permanent residence. The 501c(3) corporation will reinforce Florida's standing as the most veteran-friendly state in the nation by helping veterans to bring or develop the skills needed by Florida businesses to the state. The board of directors is comprised of nine individuals from across the state. For more information, visitwww.veteransflorida.org.

Transformación 2. Cinco países … el enigma continúa....Evolución Económica 2000-2014

El antecedente de este libro fue  el publicado en el año 2000 "Transformación, Cinco países…  un enigma" en el que se analizo la evolución económica de Argentina Brasil, Chile, México y Venezuela de 1983 a 1999.

Este nuevo libro muestra la evolución económica en el periodo 1999-2013 de esos mismo cinco países siguiendo la innovadora metodología del anterior usando diversos indicadores para observar su transformación en los últimos 14 años. En este  libro los autores, como bien afirma Don Enrique V. Iglesias en el Prologo “ han seguido una innovadora metodología en ambos períodos y ello es lo que hace particularmente interesante la empresa que se han propuesto”.

En el libro anterior se presentaban las razones del atraso que en el periodo 1983-1999 registro  Venezuela en transformación y progreso frente a los países mencionados. En este nuevo periodo el Gobierno conto con más recursos y mayor poder que en periodo anterior pero en lugar de progreso en la transformación económica  hay un retroceso en los principales indicadores, claramente responsabilidad de la ineficiencia del  modelo económico populista de control gubernamental de la economía y del Capitalismo de Estado, del llamado  eufemísticamente  “Socialismo del Siglo XX”, una disfrazada imitación del Socialismo que solo sobrevive en Cuba y Corea del Norte cuyos indicadores económico son la evidencia del fracaso del modelo.  .

Tres datos reflejan claramente  el deterioro  económico: reducción del numero de empresas, emigración  calificada y devaluación. De las 600.000 empresas operativas que existían en 1999 según el Instituto Nacional de Estadística 107.000 han desaparecido, una reduccion del 18%,  y en la industria pasaron de 11.000 a 7.000 en 2011^[1],  una reduccion del 36%. Se estima que alrededor de 1 millón 200 mil venezolanos han abandonado el país en busca de mejores condiciones de vida”^[2] y según una investigación del profesor Iván De La Vega, 48% de los emigrantes son calificados;  la tasa de venezolanos sin trabajo en EEUU es solo de 8%. La devaluación es extraoridnaria  pues en el mercado paralelo, para comprar un $US dollar se necesitan hoy 1230 veces mas bolivares que en el 2000.

Un libro narra con claridad los aspectos con impacto económico más resaltantes del periodo de 15 años, documentando y analizado los hechos que desembocaron en la critica situación económica de 2013, luego del periodo más largo de altos precios del petróleo disfrutado por Venezuela. 

---

[1] Economía | Dinero Venezuela sin Chávez 12 de mayo de 2012

[2] El Universal. 21 de julio 2014.

Do You Really Need a $10,000 Toilet?

1 – Toto Neorest 750H

You may not “need” a $10,000 toilet, but once you’ve experienced a Toto Neorest 750H, you will aspire to own one. I spent the week at the Kitchen & Bath Industry Show (KBIS), which is presented in conjunction with the National Kitchen and Bath Association (NKBA), and it was awesome! The $31 billion kitchen and bath industry brings together the world’s best designers, craftspeople and technologists in ways that truly impact our daily lives – up to and including high-tech toilets that keep your bottom warm and clean and practically clean themselves. Here are six things that got my undivided attention.

Yes, that’s me staring into a $10,000 toilet. All kidding aside, the engineers at Toto have taken toilet technology to a whole new level. The Neorest 750H has several remarkable features. There’s an integrated UV light in the lid that combines with a special glaze to help break down organic material. It automatically rinses the bowl with antibacterial electrolyzed water, and the results are an almost self-cleaning toilet. Then, there are the fun features like the heated seat, deodorizer and warm water sprays – which require a detailed explanation that I will not provide here.

2 – 3D-Printed Faucets from DXV by American Standard

I’m a big fan of 3D printing (additive manufacturing) and the practically infinite manufacturing opportunities it empowers. You can 3D print in almost any material. Designer Jean-Jacques L’Henaff from LIXIL Water Technology has teamed up with American Standard to demonstrate the power and potential of the art form. Today, this is an objet d’art, the only advantage is the 3D-centric design. But soon, this technology will be pervasive and it is likely have a dramatic impact on the way we shop (and manufacture) kitchen and bath hardware. While these 3D-printed faucets cost approximately $19,000 in stainless steel, for $50 more you can print them in titanium. Who wouldn’t want a 3D-printed titanium faucet for $19,050?

3 – Kalamazoo Hybrid Fire Grill

If you are a true grill master or just a weekend family griller, you will lose your mind when you see the Kalamazoo Hybrid Fire Grill in action. It allows you to use any combination of gas, wood and charcoal to create the perfect fire at the perfect temperature for anything you might want to grill. The wood/charcoal drawer makes refueling easy, and the grill is engineered to be very easy to clean. I am a particular fan of the different types of grates (meat, fish, veggies) that are available. I was hugely impressed by the entire line of Kalamazoo Outdoor Gourmet’s products, but the Hybrid Fire Grill just made me smile.

4 – Monogram Indoor Pizza Oven

When you grow up in New York, pizza is a staple food. Over the years, gourmet pizza recipes have come into their own, and nowadays you can put anything on a pizza crust and call it a pizza. Purists and adventurers alike will instantly fall in love with the Monogram Indoor Pizza Oven. This wall-mounted indoor oven takes only 25 minutes to reach 725 degrees Fahrenheit (the correct temperature for gourmet pizza). Then, you’re two minutes away from all the amazing pizza combinations you can concoct. You may be put off by the $9,900 price tag – but great pizza is worth it!

5 – MoistureShield® Composite Decking

I don’t think I was prepared for the state-of-the art in composite decking material, and I was beyond impressed by what some of the world’s best designers were capable of doing with it. In practice, you can use composite decking materials just as you would use wood. MoistureShield is a green product using 95 percent recycled material, and the company’s non-capped products look and feel better than almost everything else I’ve seen. While MoistureShield has competition from Trex®, TimberTech®, EverGrain® and Fiberon®, MoistureShield’s price, performance and warranty make it stand out.

6 – Boral Versetta Stone®

Cultured stone veneers have been around for a long time. Most products require extremely talented professionals to install them, and as far as aesthetics are concerned, only you can judge – it’s a matter of personal taste. What caught my attention about Boral Versetta Stone were the unique form factor, the mounting system and how incredibly easy it is to install. It’s packaged in 17-lb. sheets that can be installed with mechanical fasteners. Basically, anyone with a level, a hammer and a sense of what a stud is can do this by themselves. If you have a project (indoor or outdoor) that needs this kind of “stone mason” look and feel, Boral Versetta Stone is worth considering.

The Social Security Administration (SSA) is looking for two executives

The Social Security Administration (SSA) is looking for two executives to lead its Office of Income Security Programs (OISP) as the Associate Commissioner (AC) and Deputy Associate Commissioner (DAC) responsible for Retirement and Survivors Insurance (RSI) and Supplemental Security Income (SSI) policy.  The AC and DAC provide executive direction to an organization with approximately 125 employees and are responsible for ensuring programs under its purview are administered in an effective and efficient manner and align with the SSA mission and strategic objectives.  The AC will report directly to the Deputy Commissioner and Assistant Deputy Commissioner for Retirement and Disability Policy, the DAC will report directly to the AC; both positions are located at the agency’s Headquarters campus in Woodlawn, Maryland.

The positions carry responsibility for a wide range of Social Security and SSI policies regarding eligibility and benefit amounts, such as earnings and coverage requirements, defining marriage and other family relationships, SSA’s role in Medicare, and implementing provisions of the recent Bipartisan Budget Act. Specific topics include: Social Security numbers, applications and appeals, due process, and administrative finality; wages, coverage and exceptions, self-employment income, retirement earnings test, State and local coverage, windfall elimination provisions and government pension offset; SSI income, resources, and in-kind support and maintenance; preventing/reducing improper payments; representative payees; and Agency notice improvement activities.

The applicants should have a broad knowledge of Social Security laws, regulations, and policies, senior level experience developing, analyzing and/or interpreting public policy for social insurance and income maintenance programs in the public or private sector, and demonstrated senior level experience in planning, directing, managing and overseeing a large, fast-paced organization.  Activities also include human resources management, collaboration with multiple internal and external customers, knowledge of congressional legislation, assessing political and operational impact of decisions with a high degree of quality, and providing first class service to the public being served.

The complete vacancy announcement can be found on USAJOBS; direct link below.  Interested applicants may apply to either or both of the vacancies:

Associate Commissioner                   SSA-EX-463  

Deputy Associate Commissioner       SSA-EX-464  

Applicants should address the leadership and technical requirements of the position, describe work accomplishments, how they have exercised leadership to deliver significant results, experience collaborating/communicating and cooperating to achieve goals, and experience leading strategic change and overcoming obstacles to effectuate those changes. 

J. Jioni Palmer

Associate Commissioner for External Affairs

(T) 410-965-1804

@SSAOutreach

Cisco's 2016 Security Report: Attacks getting stronger, defender confidence dropping

On Tuesday, Cisco recently released the 2016 version of Annual Security Report on cybersecurity trends and challenges. Here are the highlights.

On Tuesday, January 19, Cisco released its 2016 Annual Security Report, highlighting the progression of cybersecurity and what businesses can expect as they move more into the new year. The full report (with appendix) is 87 pages long, so we'll give you the highlights.

One of the top findings from this year's report was that defender confidence is dropping, with only 45% of global organizations worldwide confident in their security relative to today's threats. However, many executive said they expect greater transparency on security in the future. According to a company press release: "This points to security as a growing boardroom concern."

Still, these growing concerns are acting as a catalyst for organizations to improve their security practices, as they know where their weaknesses are and what they need to work on.

SEE: Hackers' modus operandi: 5 insights that may help identify emerging threats

Aging infrastructure also played a role in poor security posture with 92% of internet devices operating with known vulnerabilities. Jason Brvenik, principal engineer for the Security Business Group at Cisco, said that some were running with up to 26 vulnerabilities. Additionally, 31% of devices are running with no vendor support.

"The second highest barrier to adopting advanced security practices and technologies are compatibility issues," Brvenik said.

Cisco's report also identified another, relational threat to enterprises—SMBs. Based on the report's finding, SMBs use fewer tools to identify and defend against security threats. These "structural weaknesses" present a potential risk to enterprises that may be working with SMBs in some capacity.

However, SMBs are improving their security due, in part, to outsourcing security services. All in all, outsourcing security is on the rise across the board with more than half of all larger organizations outsourcing consulting services, as well as a good number of businesses outsourcing auditing, monitoring, incident response, and more.

With all the changes in the way security is handled on the business side of things, it begs the question of where the major threats are happening now and where they'll come from in the future.

For one, social media platforms are growing as a foundation for criminals to carry out their campaigns—especially when it comes to compromised servers, like those for WordPress. Between February and October 2015, the number of WordPress domains that were being used by cybercriminals grew by a staggering 221%, according to the report.

Another growing risk is the browser—specifically, malicious browser extensions that have impacted more than 85% in terms of data leakage. Craig Williams, senior technical leader at Cisco, said that he wasn't surprised by this number, though.

"The fact of the matter is, these days, if you're not patching your browser and if you're not patching the plugins, you're going to be attacked by a massive number of threats," Williams said.

However, Williams said, there are legitimate reasons to not patch a browser, such as applications that require certain library versions. But, in this day and age, there are so many options for web browsers that there isn't really an excuse.

Being that most cybersecurity issues involve the internet, of course Cisco had to take a look at DNS risks. Of "known bad" malware, the Cisco report found that almost 92% used DNS to carry out their campaigns. While he wasn't totally surprised by the number, Williams said he would have originally guessed it to be closer to 85%.

HTTPS encrypted traffic is growing and, based on what Cisco observed in 2015, they believe it will soon become the leading form of traffic online. And, while that may seem like a good thing on the surface, it could introduce other problems.

"Although encryption can help protect consumers, it also can undermine the effectiveness of security products, making it more difficult for the security community to track threats," the report said. "Adding to the challenge, some malware may initiate encrypted communications across a diverse set of ports."

The three big takeaways

• Attacks are increasing and organization are losing confidence in their ability to stop them, which could serve as a catalyst for greater investments in security and greater demand for third party and cloud security services.
• SMBs have particular risks, and so do the larger organizations that partner with them.
• The rise of HTTPS to secure web traffic offers new protections, but it's not a silver bullet and could be co-opted by attackers to better cover their tracks.

The Unreasonable Reputation of Neural Networks (Shelly Palmer)

January 12, 2016

It is hard not to be enamoured by deep learning nowadays, watching neural networks show off their endless accumulation of new tricks. There are, as I see it, at least two good reasons to be impressed:

(1) Neural networks can learn to model many natural functions well, from weak priors.
The idea of marrying hierarchical, distributed representations with fast, GPU-optimised gradient calculations has turned out to be very powerful indeed. The early days of neural networks saw problems with local optima, but the ability to train deeper networks has solved this and allowed backpropagation to shine through. After baking in a small amount of domain knowledge through simple architectural decisions, deep learning practitioners now find themselves with a powerful class of parameterised functions and a practical way to optimise them.

The first such architectural decisions were the use of either convolutions or recurrent structure, to imbue models with spatial and temporal invariances. From this alone, neural networks excelled in image classification, speech recognition, machine translation, atari games, and many more domains. More recently, mechanisms for top-down attention over inputs have shown their worth in image and natural language tasks, while differentiable memory modules such as tapes and stacks have even enabled networks to learn the rules of simple algorithms from only input-output pairs.

(2) Neural networks can learn surprisingly useful representations
While the community still waits eagerly for unsupervised learning to bear fruit, deep supervised learning has shown an impressive aptitude for building generalisable and interpretable features. That is to say, the features learned when a neural network is trained to predict P(y|x) often turn out to be both semantically interpretable and useful for modelling some other related function P(z|x).

As just a few examples of this:

• Units of a convolutional network trained to classify scenes often learn to detect specific objects in those scenes (such as a lighthouse), even though they were not explicitly trained to do so (Zhou et al., 2015)
• Correlations in the bottom layers of an image classification network provide a surprisingly good signature for the artistic style of an image, allowing new images to be synthesised using the content of one and the style of another (Gatys et al., 2015)
• A recurrent network [correction below] taught to predict missing words from sentences learns meaningful word embeddings, where simple vector arithmetic can be used to find semantic analogies. For example:
  • v_king - v_man + v_woman ≈ v_queen
  • v_Paris - v_France + v_Italy ≈ v_Rome
  • v_Windows - v_Microsoft + v_Google ≈ v_Android

I have no doubt that the next few years will see neural networks turn their attention to yet more tasks, integrate themselves more deeply into industry, and continue to impress researchers with new superpowers. This is all well justified, and I have no intention to belittle the current and future impact of deep learning; however, the optimism about the just what these models can achieve in terms of intelligence has been worryingly reminiscent of the 1960s.

Extrapolating from the last few years’ progress, it is enticing to believe that Deep Artificial General Intelligence is just around the corner and just a few more architectural tricks, bigger data sets and faster computing power are required to take us there. I feel that there are a couple of solid reasons to be much more skeptical.

To begin with, it is a bad idea to intuit how broadly intelligent a machine must be, or have the capacity to be, based solely on a single task. The checkers-playing machines of the 1950s amazed researchers and many considered these a huge leap towards human-level reasoning, yet we now appreciate that achieving human or superhuman performance in this game is far easier than achieving human-level general intelligence. In fact, even the best humans can easily be defeated by a search algorithm with simple heuristics. The development of such an algorithm probably does not advance the long term goals of machine intelligence, despite the exciting intelligent-seeming behaviour it gives rise to, and the same could be said of much other work in artificial intelligence such as the expert systems of the 1980s. Human or superhuman performance in one task is not necessarily a stepping-stone towards near-human performance across most tasks.

By the same token, the ability of neural networks to learn interpretable word embeddings, say, does not remotely suggest that they are the right kind of tool for a human-level understanding of the world. It is impressive and surprising that these general-purpose, statistical models can learn meaningful relations from text alone, without any richer perception of the world, but this may speak much more about the unexpected ease of the task itself than it does about the capacity of the models. Just as checkers can be won through tree-search, so too can many semantic relations be learned from text statistics. Both produce impressiveintelligent-seeming behaviour, but neither necessarily pave the way towards true machine intelligence.

I’d like to reflect on specifically what neural networks are good at, and how this relates to human intelligence. Deep learning has produced amazing discriminative models, generative models and feature extractors, but common to all of these is the use of a very large training dataset. Its place in the world is as a powerful tool for general-purpose pattern recognition, in situations where n and d are high. Very possibly it is the best tool for working in this paradigm.

This is a very good fit for one particular class of problems that the brain solves: finding good representations to describe the constant and enormous flood of sensory data it receives. Before any sense can be made of the environment, the visual and auditory systems need to fold, stretch and twist this data from raw pixels and waves into a form that better captures the complex statistical regularities in the signal*. Whether this is learned from scratch or handed down as a gift from evolution, the brain solves this problem adeptly - and there is even recent evidence that the representations it finds are not too dissimilar from those discovered by a neural network. I contend, deep learning may well provide a fantastic starting point for many problems in perception.

That said, this high n, high d paradigm is a very particular one, and is not the right environment to describe a great deal of intelligent behaviour. The many facets of human thought include planning towards novel goals, inferring others' goals from their actions, learning structured theories to describe the rules of the world, inventing experiments to test those theories, and learning to recognise new object kinds from just one example. Very often they involve principled inference under uncertainty from few observations. For all the accomplishments of neural networks, it must be said that they have only ever proven their worth at tasks fundamentally different from those above. If they have succeeded in anything superficially similar, it has been because they saw many hundreds of times more examples than any human ever needed to.

Deep learning has brought us one branch higher up the tree towards machine intelligence and a wealth of different fruit is now hanging within our grasp. While the ability to learn good features in high dimensional spaces from weak priors with lots of data is both new and exciting, we should not fall into the trap of thinking that most of the problems an intelligent agent faces can be solved in this way. Gradient descent in neural networks may well play a big part in helping to build the components of thinking machines, but it is not, itself, the stuff of thought.