Cloud Hosting

Business analytics is no longer a priority for tech companies alone. In 2022, McKinsey estimated that it could create approximately $1.2 trillion a year in value for government entities. A recent EY study found that 93% of companies said they plan to increase investments in the area of data and analytics.

People realize thats the way the world is going, says Daniel Guetta, an associate professor and director of the Business Analytics Initiative at Columbia Business Schooland if professionals want to succeed in their jobs, they need to have at least a basic comprehension of how data functions and how it can be leveraged in decision-making.

ADVERTISEMENT

Learn how to analyze, interpret, and visualize data for a career. Our course will prepare you to organize and interpret data professionally with hands-on projects and 1-on-1 mentoring. Gain experience in the critical tools for todays data analysts: Python, SQL, Tableau, and Excel.

With demand for this expertise growing across sectors and the technology surrounding it becoming more convoluted every day, theres incredible value in having a degree in business analytics. Companies are also paying top dollar for this expertise, with Glassdoor reporting that business analysts with MBAs make on average around $104,000 annually.

Heres what an MBA in business analytics entails, and why it might be a degree worth considering.

Business analytics is the use of technical tools and frameworks to mine data for insights that can drive value for an organization. Guetta likes to think of the field as a spectrum.

On one end, you have data analytics, data science, and business intelligence, which, while highly dependent on the company and context, generally involve collecting, storing, and then sorting through data to make informed decisions around marketing, sales, operations, or other areas of the business. I have a big data setits too big for Excel, or maybe its not too big for Excel, but its kind of complexand I just want to ask very simple questions out of it, Guetta says.

On the other end, you might have more complex concepts like predictive analytics and machine learning that focus less on comprehending the present and more on forecasting and preempting whats to come. All points in that spectrum are incredibly useful, incredibly important in different ways, he adds.

Big dataa large, diverse data set thats often difficult to process with simple computational toolsplays a role in many of these categories, particularly in emerging tech like AI that cant operate without a complicated array of inputs.

While some analytics skills can be taught on the job, a more formal education will provide you with the credibility and confidence to incorporate data into your workflows. And its not just for those who are good at math and engineers: I cannot think of someone today in 2024 who cant at least benefit in some way from knowing a little bit about coding, Guetta says, adding that his students range from aspiring product managers to venture capitalists.

I can think of at least three or four students in the last two years who work in museums whove decided they wanted to take this class and just understand how this works and whats happening and where its going.

Peter Fader, a professor of marketing at The Wharton School of the University of Pennsylvania whos taught the topic for more than three decades, agrees that anyone could benefit from an MBA in business analytics. Youve been in an organization, two, or three, and realized all the dysfunction and opportunity to improvea well-taken MBA can be immensely helpful, he says.

Specializations like business analytics used to really only be reserved for masters or PhD programs. But business schools are beginning to recognize that they must incorporate data and engineering concepts into their curriculum if they want their students to be competitive in todays job market.

The difference between a masters in business analytics and an MBA in business analytics comes down to whats taught and how deep the lesson plans go. The MS programs, depending on where and how theyre taught, are going to be a whole level way beyond the analytics courses we see in the MBA, Fader says.

In a business analytics MBA program, you might take some courses on statistics, data analysis, and emerging tech like AI, but youll also spend as much time, if not more, on business foundations such as accounting, people management, and supply chain. An MS in business analytics, meanwhile, may include classes where you get really technical and mathematical.

At Columbia Business School, Guetta says, the business analytics speciality can be broken up into three buckets. The first is fundamentals, where you learn about whats going on under the hood of specific technologies and ideas. The second is tools, where you learn how to deploy software and programming languages like SQL. And the third is applications, where youre taught to apply business analytics knowledge and tools to marketing, real estate, sports, and other areas.

Masters degrees also tend to be a better fit for recent college graduates or young professionals, while an MBA may make more sense for someone whos already fairly established in their career and/or is looking to make a major switch. Most MBA programs are going to require a few years of experience after your undergrad, Guetta adds.

Another option is a dual MS/MBA, which Columbia, along with other legacy institutions like Harvard and Emory, offer. You get the best of all worldsyou get the very rigorous engineering training, and you also get the MBA training, Guetta says.

If an MBA in business analytics is piquing your interest, consider the following (not at all exhaustive) list of institutions with in-person, virtual, and/or hybrid offerings. Note, too, that some business schools call it a masters instead of an MBA, while others provide concentrations and certifications in business analytics through their MBA programs.

Whether youre an entrepreneur, a finance professional, or a nonprofit worker, having a business analytics background can help you make smart, data-informed choices for your company or team. It will not only make you an easily marketable candidate for job openings or promotions, but it will ensure youre staying on top of the latest innovations and tech. An MBA program is an ideal place for those who want to mix business acumen with engineering, while a masters degree will help you further hone your technical skill set.

Whatever route you decide, Fader recommends researching and asking around about a schools job placement success rates and services. Having that kind of transparency and putting the conversation in the hands of the students and alums themselvesand not letting the institution filter itwould be really, really helpful, he says. Guetta suggests finding an MBA cohort that provides a broad range of business analytics courses, or even partners closely with an engineering school. Lastly, exercise muscles that help you learn quicker. Business analytics, he says, has changed unrecognizably in the last two, three years, and its going to change unrecognizably in the next two, three years. The ability to pick up things as theyre thrown at you, to pick up new topics, to pick up new trends, is going to be invaluable in the field.

Continued here:

MBA in business analytics: Big data is a big deal - Fortune

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More

Briefly Bio, a startup based in London, has announced a small but meaningful $1.2 million round from Compound VC and others with the goal of making scientific experiments and data more reproducible.

In fact, so many thousands of results of scientific experiments havent yet or possibly cant be reproduced that some researchers have deemed this a reproducibility or replication crisis.

In an effort to help solve this problem, Briefly Bios platform uses large language models (LLM) like the kind powering leading AI products such as ChatGPT to turn complex lab documentation into a consistent, structured format. The idea is that anyone can use and build on the original documents in their own lab more easily.

The company has already roped in early customers for the paid version of the offering and has plans in place to create a public version of the platform for open sharing and collaboration on scientific data that it says will be similar to GitHubs code repository for sharing open source and permissively licensed software.

As GitHub helped software engineers collaborate and build on each others code, we think Briefly can help scientists and engineers do the same with their experiments, Harry Rickerby, the CEO and co-founder of Briefly Bio, said in a statement.

When scientists try to solve a complex biological problem, they take different approaches. Some methods work, some partially do the job and some dont at all, but in all cases, the protocol for the lab work the plan for research, covering objectives, design, methodology and statistics and the details of the experiment itself are documented thoroughly.

The idea behind collating this data is to give other scientific teams a base of sorts to continue the research or solve any other closely related problem. However, this is also where the problem of reproducibility begins.

Essentially, every scientist has their own way of documenting their work, which in many cases leads to ambiguity and the loss of critical details crucial for shared understanding.

For instance, some researchers may go into extensive detail when describing their approach to gene editing in their own words, while others may just scratch the surface with the notion that other teams may have similar knowledge. This can easily lead to inefficient collaboration and failure at reproducing experiments, costing the industry over $50 billion each year.

Rickerby told VentureBeat he and his colleagues at LabGenius Katya Putintseva and Staffan Piledahl saw the problem first-hand at different levels.

Katya worked as a scientist in academia and faced the challenges of re-using and adapting work from the published literature. She then moved into data science, where she needed to understand precisely how the data was generated to analyze and model it. On the other hand, Staffan worked as an automation engineer and needed complete definitions of a lab workflow to transfer them to robots. After leaving, we realized that many struggles across our careers shared a common root cause there wasnt consistent documentation of how lab work was being run, he said.

To address this, the trio came together and launched Briefly Bio. At the core, the company provides scientists with a platform that can convert any scientific protocol documented in natural language into a consistent, structured format containing step-by-step information. All the user has to do is provide the blob of text from the original author and the tool comes up with a structured output detailing the method for reproducing or building on the experiment.

Brieflys tool is powered by generative AI, which helps structure plain text descriptions of procedural knowledge and convert them into a hierarchical representation. The large language models under the hood automatically extract the key pieces of information and categorize them into different processes, actions, explanations and parameters. This structured representation is then transformed into a visual representation that is clearer and easier to digest than a wall of text, Rickerby explained.

The offering not only creates a shared language for data understanding but also paints a clear picture showcasing how scientific methods change and evolve, in a way that just hasnt been possible with traditional text descriptions.

More importantly, in addition to converting existing scientific descriptions into a structured format, Briefly Bio also includes an AI copilot, which can be triggered via natural language to spot errors and find as many parameters as it can find in connection to the lab work being done. The AI generates missing parameters in a matter of seconds, enriching the hierarchical representation of the method for reuse in a lab experiment.

The CEO did not share the exact details of the models powering the whole experience but said they are building on top of existing models, enriching them with additional experimental context to improve the lab work understanding.

For reusing the generated data in experiments, teams can launch Briefly Bios workspace. It copies the enriched, structured method as is while allowing users and their team members to mark each step as complete/incomplete with associated calculations, text and sample layout painting a picture of whats in each well of the users plate, layer by layer.

While Briefly Bio is still in its early stages, the company claims it has started booking revenue from first customers on a per-user-based SaaS model.

Our users are typically wet lab scientists, working in early-stage research and development whether this is in academia, or in biotech and pharma and looking for a clearer way to document and share their work. Weve also found a lot of interest from those working in laboratory automation, using Briefly as a way to collaborate with scientists to properly describe their workflows before they program the robots, Rickerby noted.

In the long run, the company wants to build this work and also open up a public version of the platform for sharing experiments and protocols. This will allow scientists to discover complete, reproducible methodology that they can easily adapt and use in their own labs just as Github did for open-source software development, the CEO added.

VB Daily

Stay in the know! Get the latest news in your inbox daily

By subscribing, you agree to VentureBeat's Terms of Service.

Thanks for subscribing. Check out more VB newsletters here.

An error occured.

The rest is here:

Briefly Bio raises $1.2M to build the GitHub of science experiments - VentureBeat

Visualising AI project launched by Google DeepMind. From Unsplash image.

Gemini Pro can handle an astonishing 2M token context compared to the paltry 15k we were amazed by when GPT-3.5 landed. Does that mean we no longer care about retrieval or RAG systems? Based on Needle-in-a-Haystack benchmarks, the answer is that while the need is diminishing, especially for Gemini models, advanced retrieval techniques still significantly improve performance for most LLMs. Benchmarking results show that long context models perform well at surfacing specific insights. However, they struggle when a citation is required. That makes retrieval techniques especially important for use cases where citation quality is important (think law, journalism, and medical applications among others). These tend to be higher-value applications where lacking a citation makes the initial insight much less useful. Additionally, while the cost of long context models will likely decrease, augmenting shorter content window models with retrievers can be a cost-effective and lower latency path to serve the same use cases. Its safe to say that RAG and retrieval will stick around a while longer but maybe you wont get much bang for your buck implementing a naive RAG system.

Advanced RAG covers a range of techniques but broadly they fall under the umbrella of pre-retrieval query rewriting and post-retrieval re-ranking. Lets dive in and learn something about each of them.

Q: What is the meaning of life?

A: 42

Question and answer asymmetry is a huge issue in RAG systems. A typical approach to simpler RAG systems is to compare the cosine similarity of the query and document embedding. This works when the question is nearly restated in the answer, Whats Meghans favorite animal?, Meghans favorite animal is the giraffe., but we are rarely that lucky.

Here are a few techniques that can overcome this:

The nomenclature Rewrite-Retrieve-Read originated from a paper from the Microsoft Azure team in 2023 (although given how intuitive the technique is it had been used for a while). In this study, an LLM would rewrite a user query into a search engine-optimized query before fetching relevant context to answer the question.

The key example was how this query, What profession do Nicholas Ray and Elia Kazan have in common? should be broken down into two queries, Nicholas Ray profession and Elia Kazan profession. This allows for better results because its unlikely that a single document would contain the answer to both questions. By splitting the query into two the retriever can more effectively retrieve relevant documents.

Rewriting can also help overcome issues that arise from distracted prompting. Or instances where the user query has mixed concepts in their prompt and taking an embedding directly would result in nonsense. For example, Great, thanks for telling me who the Prime Minister of the UK is. Now tell me who the President of France is would be rewritten like current French president. This can help make your application more robust to a wider range of users as some will think a lot about how to optimally word their prompts, while others might have different norms.

In query expansion with LLMs, the initial query can be rewritten into multiple reworded questions or decomposed into subquestions. Ideally, by expanding the query into several options, the chances of lexical overlap increase between the initial query and the correct document in your storage component.

Query expansion is a concept that predates the widespread usage of LLMs. Pseudo Relevance Feedback (PRF) is a technique that inspired some LLM researchers. In PRF, the top-ranked documents from an initial search to identify and weight new query terms. With LLMs, we rely on the creative and generative capabilities of the model to find new query terms. This is beneficial because LLMs are not restricted to the initial set of documents and can generate expansion terms not covered by traditional methods.

Corpus-Steered Query Expansion (CSQE) is a method that marries the traditional PRF approach with the LLMs generative capabilities. The initially retrieved documents are fed back to the LLM to generate new query terms for the search. This technique can be especially performant for queries for which LLMs lacks subject knowledge.

There are limitations to both LLM-based query expansion and its predecessors like PRF. The most glaring of which is the assumption that the LLM generated terms are relevant or that the top ranked results are relevant. God forbid I am trying to find information about the Australian journalist Harry Potter instead of the famous boy wizard. Both techniques would further pull my query away from the less popular query subject to the more popular one making edge case queries less effective.

Another way to reduce the asymmetry between questions and documents is to index documents with a set of LLM-generated hypothetical questions. For a given document, the LLM can generate questions that could be answered by the document. Then during the retrieval step, the users query embedding is compared to the hypothetical question embeddings versus the document embeddings.

This means that we dont need to embed the original document chunk, instead, we can assign the chunk a document ID and store that as metadata on the hypothetical question document. Generating a document ID means there is much less overhead when mapping many questions to one document.

The clear downside to this approach is your system will be limited by the creativity and volume of questions you store.

HyDE is the opposite of Hypothetical Query Indexes. Instead of generating hypothetical questions, the LLM is asked to generate a hypothetical document that could answer the question, and the embedding of that generated document is used to search against the real documents. The real document is then used to generate the response. This method showed strong improvements over other contemporary retriever methods when it was first introduced in 2022.

We use this concept at Dune for our natural language to SQL product. By rewriting user prompts as a possible caption or title for a chart that would answer the question, we are better able to retrieve SQL queries that can serve as context for the LLM to write a new query.

See the original post here:

Advanced Retrieval Techniques in a World of 2M Token Context Windows, Part 1 | by Meghan Heintz | Jul, 2024 - Towards Data Science