koki1983 wrote: ...evo,i službeno je-CS potvrdio upravo:
Crowdflower will become Figure Eight
30 Mar 2018 09:47 am
Members, on Tuesday April 3rd CrowdFlower will be officially renamed to Figure Eight. Nothing is really going to change other than the name. You will continue to access tasks as you have in the past and there should be no interruption of service once this change takes place.
We look forward to what the future holds for Figure Eight and ClixSense.
The ClixSense Team
(...-nadam se da to znači i više posla i bolje plačeni,time will tell).
Evo i malo vise informacija (doduse na engleskom jeziku) o ovoj promeni:
"Focused on the Future with a New NameFebruary 27, 2018
Today, I want to share an important company decision. In the upcoming weeks, we will be changing our company name from CrowdFlower to Figure Eight. Nothing is changing today, but work is already underway and the transition will be complete in Q2.
Changing a company name is an emotional topic because it’s about identity. To the outside world, a corporate name change often appears as an abrupt, out of the blue announcement catching customers, partners, and other constituents off guard.
The reality is that something this fundamental is usually years in the making. In our case, it’s the culmination of a shift in our business, advancement in our technology, and significant maturation of our customers and industry over the past three years.
The Founding of CrowdFlower
Search was the first large scale application of AI in our daily lives and our founders Lukas Biewald and Chris Van Pelt came from the world of search (Yahoo and Powerset). While trying to make AI work in search, they realized that the bottleneck to success was not tuning the algorithm, but rather collecting large amounts of high quality, human-labelled training data.
But obtaining this training data was not easy. At the time you had two horrible options. Horrible option number 1 was to ask your engineering team to stop coding and start labeling data. This was a highly expensive misallocation of resources. Horrible option number 2 was to give it to some third party and hope the data you got back was high quality, but you lost control of the process and the quality was rarely good enough. So Lukas and Chris decided to found CrowdFlower to find a better third option that met the requirements of quality, cost, flexibility, and control. They built layers of quality control on top of a community of people willing to do microtasks for a fee, so the name CrowdFlower made sense. The quest of using humans to better train AI algorithms has been embedded in our company’s DNA from Day One.
The Evolution of the AI and Machine Learning Market
Fast forward to today and the adoption of AI and Machine Learning within the enterprise has changed dramatically, both quantitatively and qualitatively. At CrowdFlower, we’ve been one of the companies fortunate enough to be a contributor to the adoption of AI and we’ve seen an acceleration in how AI is being adopted over the last 3 years, including:
Adoption in every industry: Three years ago AI adoption was really only concentrated in a few industries such as the retail and technology verticals. Today every vertical – from Financial Services to Healthcare to Government to Manufacturing to Consumer Goods to Automotive to Media – is working on applying machine learning to their core business processes.
Training data usage: Three years ago, the typical usage pattern for training data was “one and done” batches of 100,000 labels where models were not being deployed to production. Now increasingly we see training data volume of 1,000,000+ labels being continuously integrated into production machine learning models.
Budgets for training data: Three years ago there was not the widespread understanding that training data was a new important asset that needed its own budget. Now machine learning teams know they have to budget for training data as seriously as they have to budget for their technology infrastructure.
Complexity of labeling tasks: Three years ago the dominant data type being labeled was text, with some image categorization. Now we’re seeing an explosion in pixel level semantic segmentation image annotation, video, LIDAR, and audio.
Data science and machine learning Expertise: Three years ago only a few hundred companies had employees with the titles of Data Scientist or Machine Learning Engineers. Now, tens of thousands of companies have at least one Data Scientist or Machine Learning Engineer, and over 500 universities offer Data Science degrees. Even still, there is a scarcity of the necessary expertise.
Machine learning platforms: Three years ago, if you wanted a machine learning stack you had to build your own. Now machine learning teams have a choice of platforms from Amazon, Google, IBM, Intel, Microsoft, NVIDIA, Oracle, and Salesforce, or open source frameworks such as TensorFlow and PyTorch.
Our Adaptation to Serve Our Customers
As the market has evolved, CrowdFlower has adapted to serve our customers and their changing needs. We’ve executed intentional changes to how we operate to meet the market requirements of scale, quality, complexity, and flexibility. Some of the major changes include:
Business model shift from managed service to SaaS: Three years ago CrowdFlower was a managed service where customers were dependent on CrowdFlower employees to create their training data. Now CrowdFlower is a SaaS platform that customers can use directly themselves to create the large-scale, high quality customized training data sets.
Abstracted sourcing human intelligence from the application of human intelligence: CrowdFlower’s core competence is not providing human intelligence, but rather the application of human intelligence to the problem of improving machine learning models. While 3 years back, many clients came to us simply seeking the crowd of contributors we had assembled to work on data labeling tasks, now we offer a platform that has 3 choices for which categories of humans work on your data. First choice is our trusted contributors, second choice is a BPO, third choice is your own employees. Our platform ensures optimal output whatever source of human intelligence works for our clients’ projects.
Human-in-the-loop and active learning: Three years ago CrowdFlower was only used upstream from the machine learning models. Now CrowdFlower is used upstream and downstream from machine learning models. The upstream component is the initial training data, the downstream component is handling the low confidence output from the model which then closes the loop by becoming new training data upstream back into the model. These “human-in-the-loop” deployments result in active learning and the CrowdFlower platform continuously improving the output of machine learning models.
Combining human intelligence and machine learning: Three years ago the CrowdFlower solution was 100% human powered. Now, our platform uses a smart combination of human intelligence and machine learning. For example, when labeling pixels in an image, our platform will use machine intelligence to do a first pass, and then a human will course correct where it has made an error.
Acting as a trusted AI guide: Three years ago CrowdFlower served customers in a more limited way. They gave us their unstructured data, we returned their structured data, and that was the end of the relationship. Now we have customers asking us to be their trusted guide throughout their AI journey, from training data to machine learning model creation and model deployment to active learning and human-in-the-loop. We’ve created a customer-facing machine learning team to help customers throughout their journey of making AI work for their business.
There are no shortages of quotes and hokey metaphors about the bittersweet nature of growth. All reflect the same general sentiment. Moving on means eschewing a comfort we have learned to love. Our company’s time as CrowdFlower has been a wild journey of innovation, hard work by brilliant minds, partnering with customers on ground-breaking projects and playing our part in moving the AI industry forward. With nostalgia and pride and for all of the accomplishments listed above, it became clear that the time had come to move on from a name and brand we had grown to love.
Why Figure Eight?
Given the emotional rollercoaster aspect of changing a company name, we knew we needed a professional firm to help us navigate through it. So we brought in one of the top naming agencies in the valley A Hundred Monkeys to help us through a structured process that took 4 months. We gave ourselves time to explore and iterate. We interviewed employees, advisory board members, and customers. We aligned on key criteria before evaluating names. We dug deep to understand, not just who we had become, but where we would be headed. We explored the concepts of continuous learning loops, the dualities between humans and machines, and ways in which learning happens.
We had hopes that our new name would capture key elements of who we are and where we were going and had been warned by our trusted advisors at A Hundred Monkeys that a good name takes time to grow on you. We eventually chose Figure Eight because it evokes both a shape and a number that is meaningful to us.
The shape of a Figure Eight is a continuous loop you can create without your pen leaving the page in the same way the algorithm needs to continuously learn from training data and humans-in-the-loop.
The shape of a Figure Eight is also a method for training horses, a way for humans to control nature. The shape allows the rider to stay in one direction and also change direction which are both essential training steps. Horses internalize these patterns to develop autonomy much like machine learning models use training data.
The number eight is meaningful to us because a byte is a building block of 8 bits. A byte is the number of bits used to encode a single character of text (e.g. “a” or “7”) in a computer. This was a key innovation in computing to make language generated by humans readable by machines. So the number eight represents a communication path between humans and machines.
The number eight is also the difference in the atomic numbers for Carbon (6) and Silicon (14). Carbon is an essential component of human biology. Silicon is an essential component of computing power. So the number eight represents the bridge between humans and machines.
So What’s Next?
Nothing changes today. We’re still CrowdFlower for now and we won’t be making the changeover until Q2 2018. As we get closer to the changeover date, we shall share more but in the spirit of transparency and openness, we wanted to share this coming change with our community of customers, partners, and contributors sooner rather than later.
During the transition, as CEO, I have the great fortune to reflect on the CrowdFlower team and their accomplishments that have gotten us to this point. I couldn’t be prouder. I also can see into our future as Figure Eight. Like the ongoing loop our new name represents, our path ahead will be one of iteration, continuous learning, and forging the future of how humans and machines work together."